Facility-Level Mutation Fingerprints and Early-Warning Surveillance of Drug-Resistant Mycobacterium tuberculosis in Rural Eastern Cape

Facility-Level Mutation Fingerprints and Early-Warning Surveillance of Drug-Resistant Mycobacterium tuberculosis in Rural Eastern Cape | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Facility-Level Mutation Fingerprints and Early-Warning Surveillance of Drug-Resistant Mycobacterium tuberculosis in Rural Eastern Cape Lindiwe Modest Faye, Melisa Makhuba, Ntandazo Dlatu, Mojisola Clara Hosu, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8790685/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 6 You are reading this latest preprint version Abstract Background Routine molecular tuberculosis (TB) diagnostics generate high-dimensional "off-protocol" data, including mutant melt peak temperatures and cycle threshold (CT) values. These data are currently underutilized, typically discarded after individual resistance reporting. We aimed to evaluate whether aggregating these routine "mutation-proxy" signals could provide a scalable framework for facility-level surveillance and early warning of emerging drug resistance. Methods We conducted a retrospective longitudinal study of 4,300 TB diagnostic episodes across 139 health-care facilities over 16 quarters. Mutation-proxy signals for five key loci ( katG, inhA, gyrA, rrs, eis ) were extracted from raw diagnostic outputs. We constructed "facility-level mutation fingerprints" by aggregating prevalence data and employed hierarchical clustering to identify distinct resistance topographies. Associations between proxies and laboratory-confirmed resistance were modelled using L2-regularized (ridge) logistic regression with facility-level cluster bootstrap confidence intervals to account for near-separation and spatial autocorrelation. Results Isoniazid-associated proxies predominated ( katG : 46.3%; inhA : 25.1%), while gyrA (fluoroquinolone-associated) and rrs (injectable-associated) proxies were detected in 12.5% and 7.7% of episodes, respectively. Clustering revealed four distinct facility profiles: katG -dominant, inhA -dominant, mixed-isoniazid, and a high-risk "emerging gyrA " profile. Regression analysis confirmed high diagnostic accuracy for the proxies, notably for isoniazid (katG: OR = 1,146; inhA: OR = 603) and fluoroquinolones (gyrA: OR = 7,136). Longitudinal analysis successfully identified a subset of facilities that exhibited significant quarter-over-quarter increases in second-line resistance proxies prior to traditional surveillance detection. Conclusion Facility-level mutation fingerprinting leverages existing, "near-zero-cost" laboratory data to provide a granular, real-time map of the resistance landscape. This framework enables precision public health interventions, allowing TB programmes to transition from reactive to proactive, facility-targeted containment of emerging drug-resistant Mycobacterium tuberculosis . Mycobacterium tuberculosis drug resistance molecular surveillance mutation proxies facility-level analysis early-warning systems Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Tuberculosis (TB) remains a formidable public health challenge in South Africa, exacerbated by the persistent rise of drug resistance and unabated transmission within both healthcare and community settings [ 1 , 2 ]. The advent of molecular diagnostic platforms has revolutionized TB detection, enabling rapid identification of resistance-associated mutations to guide clinical decision-making [ 3 , 4 ]. However, a significant gap remains: the high-resolution molecular outputs generated by these platforms are primarily restricted to individual patient management, leaving their potential for population-level surveillance and health-system monitoring largely untapped. Routine molecular assays employed in the diagnosis of Mycobacterium tuberculosis produce more than just binary results; they generate rich, high-resolution proxy signals of genetic variation, including mutant and wild-type melt peak temperatures and cycle threshold values [ 5 , 6 , 7 ]. These signals directly reflect underlying mutations in critical resistance-associated loci, such as katG , inhA , gyrA , rrs , and eis [ 8 , 9 ]. Despite being generated at scale, these data are rarely synthesized to examine the spatial, temporal, or facility-level patterns of mutational burden and the kinetics of resistance emergence. Conventional TB drug-resistance surveillance currently relies on aggregated phenotypic outcomes or categorical molecular interpretations (e.g., "Susceptible" vs. "Resistant"). While these metrics are essential, they are inherently reactive and may obscure the subtle, early shifts in mutational landscapes that precede widespread clinical resistance. In high-burden settings, healthcare facilities and districts often exhibit distinct patient case-mixes, treatment histories, and referral patterns. These unique local dynamics impose heterogeneous selective drug pressures, driving facility-specific trajectories of resistance emergence and circulation that are invisible to standard reporting mechanisms [ 10 , 11 , 12 ]. Repurposing these routine molecular outputs for facility-level analysis represents a paradigm shift from passive reporting to proactive genomic proxy surveillance. By aggregating these signals over time and geography, we can define "mutation fingerprints" that characterize the dominant and emerging resistance profiles of specific healthcare sites. These fingerprints provide a critical window for intervention, serving as early-warning indicators that allow for targeted programmatic responses before resistance becomes entrenched within a facility or community [ 13 , 14 ]. This study, therefore, aimed to develop and evaluate a facility-level mutation fingerprinting and early-warning surveillance framework using routine molecular diagnostic data. By integrating mutation-proxy signals, resistance interpretations, and temporal trends with health-system metadata, we demonstrate how underutilized laboratory data can be transformed into actionable intelligence to strengthen TB control. 2. Methodology 2.1 Study design A retrospective analytical study was conducted using routinely collected molecular diagnostic data for Mycobacterium tuberculosis. The study applied a mixed descriptive, inferential, and exploratory modelling approach to characterize facility-level mutation patterns and assess their relationship with interpretations of drug resistance over time. 2.2 Data source and study population The dataset comprised TB diagnostic records generated through laboratory molecular diagnostic testing platforms. Each record corresponded to a patient diagnostic sample and included unique patient and identifiers, specimen collection dates, facility and geographic metadata, molecular assay outputs, and drug-resistance interpretation results. All records with valid specimen dates and facility identifiers were eligible for inclusion. Records with incomplete key identifiers were excluded from facility-level aggregation but retained for descriptive summaries where appropriate. 2.3 Variables and operational definitions 2.3.1 Facility and geographic variables Facility-level analysis was conducted using routinely recorded geographic and service-related variables. These variables were used to characterize the spatial distribution of diagnostic episodes and to support aggregation of molecular surveillance data across health system levels. For this study, health-care facilities were treated as the primary unit of surveillance analysis, enabling facility-level aggregation of mutation patterns, temporal trends, and early-warning signals. 2.3.2 Temporal variables Specimen collection dates were used to derive monthly and quarterly time periods for temporal trend and early-warning analyses. 2.3.3 Mutation-proxy variables Mutation-proxy signals were derived from mutant melt peak temperature outputs for resistance-associated loci, including katG and inhA (isoniazid-associated), gyrA regions (fluoroquinolone-associated), and rrs and eis (injectable-associated). For each locus, mutation presence was operationalized as a binary indicator, with mutant melt peak temperatures coded as 1 (detected) and absence coded as 0. Wild-type melt peak temperatures and cycle threshold values were retained as continuous variables to support the assessment of signal characteristics. 2.3.4 Drug-resistance interpretation variables Resistance interpretation variables were extracted for isoniazid (INH), fluoroquinolones (FLQ), amikacin (AMK), kanamycin (KAN), capreomycin (CAP), and ethionamide (ETH). Standard interpretation categories, namely susceptible, resistant, low-level resistant, uninterpretable, and not tested, were analyzed descriptively and incorporated as outcome variables in regression modelling. 2.4 Construction of facility mutation fingerprints Facility mutation fingerprints were constructed by aggregating mutation-proxy indicators at the facility–time level. For each facility and time, mutation prevalence was calculated as the proportion of tested episodes with a detected mutant melt peak. Facility fingerprints were represented as vectors of mutation prevalence across loci, enabling comparisons between facilities and overtime. 2.5 Exploratory clustering of facilities Unsupervised clustering techniques were applied to facility mutation fingerprints to identify groups of facilities with similar mutational profiles. Hierarchical clustering using appropriate distance measures for proportional data was employed. Cluster robustness was evaluated using internal validation metrics, and clusters were interpreted in relation to geographic location and resistance interpretation patterns. 2.6 Early-warning and temporal trend analysis To detect emerging resistance signals, time-series analyses were conducted on key mutation proxies at the facility level. Baseline prevalence estimates were established for each facility, and deviations from baseline were assessed using statistical change-detection approaches. Facilities exhibiting sustained or abrupt increases in high-risk mutation proxies were flagged as potential early-warning signals. These signals were examined alongside resistance interpretation outcomes to assess concordance and lead-time advantages over conventional reporting. 2.7 Association between mutation fingerprints and resistance outcomes Mixed-effects regression models were fitted to quantify associations between mutation-proxy patterns and drug-resistance interpretation outcomes. Facility was included as a random effect to account for clustering of episodes within facilities. Fixed effects included mutation-proxy indicators and selected temporal covariates. Model outputs were used to assess the strength and consistency of relationships between facility-level mutational burden and observed resistance patterns. 2.8 Statistical analysis All analyses were conducted using reproducible, script-based workflows. Data cleaning and preprocessing included verification of unique patient and diagnostic episode identifiers, harmonization of facility identifiers, and transformation of specimen collection dates into standardized monthly and quarterly time periods. Mutation-proxy variables were operationalized as binary indicators based on the presence or absence of mutant melt peak temperatures for resistance-associated loci ( katG , inhA , gyrA regions, rrs , and eis ). Wild-type melt peak temperatures and cycle threshold (CT) values were retained as continuous variables for descriptive assessment and sensitivity analyses. Facility-level mutation fingerprints were generated by aggregating mutation-proxy prevalence within facilities across defined time periods. Unsupervised clustering of facilities was performed using hierarchical clustering with Ward’s linkage on standardized mutation-prevalence vectors. Cluster structure and robustness were assessed using internal validation metrics and visual inspection of dendrograms. Temporal trends and early-warning signals were evaluated by analyzing facility-level mutation prevalence over time. Baseline prevalence estimates were established for each facility, and deviations from baseline were identified using change-detection approaches appropriate for proportion-based time-series data. Associations between mutation-proxy signals and drug-resistance interpretation outcomes were examined using mixed-effects logistic regression models. Resistance interpretation outcomes were treated as categorical dependent variables, with mutation-proxy indicators specified as fixed effects. Facility was included as a random intercept to account for clustering of diagnostic episodes within facilities. Model results were reported as adjusted odds ratios with corresponding 95% confidence intervals. All statistical analyses were conducted using R (version 4.3). Key packages included tidyverse for data management and visualization, lme4 for mixed-effects modelling, cluster and factoextra for clustering analyses, and changepoint for temporal change detection. Statistical significance was assessed using a two-sided alpha level of 0.05. 2.9 Sample size and power considerations This study used secondary routine laboratory data comprising approximately 4,300 Mycobacterium tuberculosis diagnostic patient samples aggregated across 139 health-care facilities. As the dataset represents near-complete routine testing during the study period rather than a sampled population, a formal a priori sample size calculation was not applicable. Nevertheless, post hoc power and precision considerations indicate that the available sample size was adequate for the planned analyses. With over 4,000 patients, the study had sufficient statistical power (> 80%) to detect small-to-moderate associations (odds ratios of approximately 1.3–1.5 or greater) between mutation-proxy indicators and drug-resistance interpretation outcomes in logistic regression models at a two-sided α level of 0.05. At the facility level, the inclusion of 139 facilities provided adequate variability to support facility-level aggregation, clustering, and mixed-effects modelling. The number of facilities exceeded recommended thresholds for stable estimation of random effects in multilevel models, allowing reliable assessment of between-facility heterogeneity in mutation patterns and resistance outcomes. For early-warning and temporal trend analyses, the large number of observations per facility over time enabled detection of meaningful deviations in mutation-proxy prevalence beyond expected random fluctuation. Facilities with low testing volumes were handled by aggregating across quarterly intervals and performing sensitivity analyses to ensure the robustness of detected signals. 3. Results 3.1 Study dataset and facility coverage A total of 4,300 tuberculosis diagnostic patients’ samples were included in the analysis, representing routine molecular testing conducted across 139 health-care facilities. All records contained valid facility identifiers and specimen collection dates, allowing for facility-level aggregation and temporal analyses. The dataset spanned 16 calendar quarters, enabling assessment of longitudinal trends and recent changes in mutation prevalence. 3.2 Prevalence of mutation-proxy signals Mutation-proxy signals were identified by the presence of mutant melt peak temperatures at resistance-associated loci. Overall, isoniazid-associated mutation proxies predominated, while fluoroquinolone- and injectable-associated proxies were less frequent. Mutant melt peaks in katG were detected in 46.3% of diagnostic patients’ samples, followed by mutant melt peaks in inhA at 25.1%. Mutation proxies in gyrA regions were identified in 12.5% of episodes, and rrs -associated mutant peaks were detected in 7.7%. No independent eis mutant peaks were observed; however, injectable-associated mutation proxies ( rrs and/or eis ) were present in 7.7% of diagnostic patients’ samples overall. The prevalence of mutation-proxy signals across all analysed diagnostic patient samples is summarized in Table 1 . Table 1 Prevalence of mutation-proxy signals detected by routine molecular testing (N = 4,300) Mutation Proxy (Gene/Locus) Number of diagnostic patient samples Percentage (%) katG mutant melt peak is present 1 991 46.3 inha mutant melt peak is present 1 080 25.1 gyrA region mutant peak present (any) 538 12.5 rrs mutant melt peak is present 331 7.7 ei s mutant melt peak present 0 0.0 Injectable-associated proxy ( rrs/eis ) 331 7.7 3.3 Drug-resistance profiles The resistance interpretation in Fig. 1 demonstrates marked heterogeneity in resistance patterns across first- and second-line anti-tuberculosis drugs. Isoniazid (INH) shows the highest burden of resistance, with 51.6% of diagnostic patients’ samples classified as resistant and an additional 17.3% as low-level resistant, indicating that nearly 7 in 10 diagnostic patients’ samples exhibit some degree of isoniazid resistance. Only 21.9% of diagnostic samples from patients were fully susceptible to INH, highlighting the central role of isoniazid resistance in our study. In contrast, fluoroquinolones (FLQ) and injectable agents (amikacin, kanamycin, and capreomycin) remained predominantly susceptible, with susceptibility proportions ranging from 76.8% to 83.5%. Resistance to these second-line drugs was comparatively uncommon, occurring in approximately 7–8% of episodes, and low-level resistance was rare or absent. These patterns suggest that while second-line resistance is present, it has not yet become widespread. Ethionamide (ETH) displayed an intermediate resistance profile, with 25.5% of diagnostic patients’ samples classified as resistant and 66.1% susceptible, consistent with known cross-resistance between ethionamide and isoniazid mediated through inhA -associated mechanisms. Across all drugs, a consistent “not tested” (NT) category of 8.1% reflects routine diagnostic workflows rather than biological resistance. Uninterpretable and uninterpretable-susceptible categories were infrequent, indicating generally robust assay performance. 3.4 Facility-level mutation fingerprints Facility-level mutation fingerprints were constructed by aggregating mutation-proxy prevalence across facilities with sufficient testing volume (≥ 30 episodes). Hierarchical clustering revealed marked heterogeneity in mutational profiles between facilities (Fig. 2 ). Some facilities were characterized by high prevalence of katG -dominant mutation patterns, while others showed inhA -dominant or mixed profiles. A smaller subset of facilities demonstrated an elevated prevalence of gyrA-associated mutation proxies, suggesting increased fluoroquinolone resistance risk. Injectable-associated mutation proxies were concentrated in a limited number of facilities. 3.5 Association between mutation proxies and drug resistance Because mutation-proxy variables exhibited strong associations with resistance interpretation outcomes, standard logistic and mixed-effects models showed evidence of near-separation, leading to unstable coefficient estimates. To address this, L2-regularised logistic regression was employed, combined with facility-level cluster bootstrap confidence intervals, which provide stable and interpretable estimates under conditions of quasi-complete separation while accounting for clustering of diagnostic episodes within facilities. 3.5.1 Isoniazid resistance For isoniazid resistance, models including katG and inhA mutation proxies demonstrated powerful associations with resistance outcomes. The presence of a katG mutant proxy was associated with an adjusted odds ratio (OR) of approximately 1,146.6 (cluster-bootstrap 95% CI: 584.0–1,928.9), while inhA mutation proxies were associated with an OR of approximately 603.6 (95% CI: 362.3–885.7) (Table 2 ). Table 2 Association between isoniazid resistance and mutation-proxy signals Mutation proxy Adjusted OR 95% CI Interpretation katG mutant proxy 1 146.6 584.0–1 928.9 Strong association with INH resistance InhA mutant proxy 603.6 362.3–885.7 Strong association with INH resistance 3.5.2 Fluoroquinolone resistance For fluoroquinolone resistance, gyrA mutation proxies were the dominant predictors, with an adjusted OR of approximately 7 136.5 (95% CI: 1 348.3–18 367.0). Secondary associations were observed for katG (OR ≈ 4.93, 95% CI: 1.09–13.78) and inhA (OR ≈ 7.16, 95% CI: 1.49–21.04), indicating that fluoroquinolone resistance frequently occurred in the context of broader resistance-associated mutation backgrounds (Table 3 ). Table 3 Association between fluoroquinolone resistance and mutation-proxy signals Mutation proxy Adjusted OR 95% CI Interpretation gyrA mutant proxy (any region) 7 136.5 1 348.3–18 367.0 Primary predictor of FLQ resistance katG mutant proxy 4.93 1.09–13.78 Secondary association inhA mutant proxy 7.16 1.49–21.04 Secondary association 3.6 Early-warning signals for emerging resistance Quarterly facility-level monitoring identified facilities with the largest recent increases in mutation-proxy prevalence between the two most recent quarters. For fluoroquinolone-associated risk, as indicated by increases in gyrA mutation proxies, the most pronounced recent increases were observed at Facility 117, Facility 126, Facility 26, Facility 69, and Facility 30 (Fig. 3 ). In parallel, analysis of injectable-associated risk, based on rrs and/or eis mutation proxies, identified Facility 9, Facility 117, Facility 21, Facility 138, and Facility 11 as facilities with the largest recent increases (Fig. 4 ). These temporal deviations are visualized using quarterly time-series plots, with dashed vertical lines marking the two most recent quarters, enabling direct comparison of recent changes relative to prior facility-specific baselines. The concentration of increases within a limited number of facilities supports the use of facility-level mutation surveillance as an early-warning mechanism to flag settings with emerging resistance-associated signals. Discussion Facility-level mutation fingerprints as indicators of local selective pressure The facility-level mutation fingerprints identified in this study reveal substantial heterogeneity in the mutational landscape of Mycobacterium tuberculosis across health-care settings. Facilities clustered into distinct groups characterized by predominance of katG -associated, inhA -associated, or mixed isoniazid-related mutation profiles, with a smaller subset exhibiting elevated prevalence of gyrA -associated mutation proxies linked to fluoroquinolone resistance. This heterogeneity indicates that resistance emergence is not uniformly distributed across the health system, but instead reflects localised selective pressures shaped by treatment practices, referral pathways, and patient case mix [ 14 , 15 , 16 ]. Likewise, WGS-based work from high-burden settings routinely identifies katG and inhA as major INH resistance loci, but with variable proportions across study populations and catchment areas [ 17 ]. A study carried out in the Western Cape, South Africa, demonstrated strong geographic clustering of ofloxacin/amikacin resistance among RR-TB, illustrating how second-line resistance can be localized rather than uniform, an epidemiologic pattern that supports our findings, facility-level gyrA subset signal [ 18 ]. Another study using whole genomic sequencing (WGS) similarly shows that fluoroquinolone resistance mutations (including gyrA) can be concentrated within specific transmission clusters, reinforcing the notion of localized emergence and spread [ 19 ]. Although our study used mutation-proxy signals (melt-peak–derived profiles) rather than WGS/LPA calls; however, the pattern of localized heterogeneity is consistent with the clustering repeatedly documented by genomic epidemiology [ 17 ]. Facilities with katG -dominant mutation profiles likely represent settings with established or recurrent isoniazid resistance, whereas inhA -dominant or mixed profiles may reflect evolving resistance pathways or differential drug exposure histories [ 20 , 21 ]. The identification of facilities with emerging gyrA -associated mutation fingerprints is of particular concern, as fluoroquinolone resistance substantially compromises the effectiveness of second-line treatment and signals progression towards more complex forms of drug-resistant TB [ 22 , 23 ]. Injectable-associated mutation proxies were concentrated in a limited number of facilities and demonstrated recent upward trends in specific settings. This is consistent with evidence that SLID resistance is largely driven by recurrent high-confidence variants (notably in rrs and eis) that can cluster within specific referral or transmission networks, even as national policy shifts toward all-oral DR-TB regimens reduce injectable use overall [ 24 , 25 ]. In a large South African genomic analysis of XDR-TB (Western Cape), SLID resistance was largely driven by a small set of recurrent variants in rrs and eis, with rrs 1401A > G being the dominant mutation pattern across lineages [ 15 ]. This clustering suggests that early emergence of second-line resistance may be driven predominantly by facility-level amplification rather than widespread community transmission. This happens especially where there is a higher concentration of patients with prior treatment, complex DR-TB, interruptions in care, or regimen changes. Large-scale analyses that distinguish baseline versus acquired resistance highlight that resistance to newer/second-line drugs can be acquired under treatment pressure, consistent with amplification dynamics [ 26 , 27 ]. In contrast, clustering does not automatically imply amplification. A study conducted in KwaZulu-Natal reported that the majority of bedaquiline-resistant cases were attributable to direct transmission of diverse resistant strains, indicating that resistance signals can become concentrated through transmission rather than through in-facility acquisition alone [ 28 ]. From a public health perspective, these findings demonstrate that routinely generated molecular diagnostic data can be repurposed to produce actionable, facility-specific surveillance intelligence, enabling targeted intervention before resistance patterns become entrenched at district or provincial levels. Facility-level mutation fingerprints provide a pragmatic bridge between individual-level diagnostics and population-level TB control, offering a scalable framework for early-warning surveillance using data already embedded within routine laboratory systems. Early-warning signals for emerging resistance Our study demonstrates the feasibility and public health value of using facility-level temporal changes in mutation proxy prevalence as early warning signals for emerging drug resistance. By focusing on quarter-to-quarter deviations rather than absolute prevalence alone, the analysis identifies facilities where resistance-associated signals are accelerating relative to their historical baselines, a pattern particularly relevant for anticipatory surveillance. The detection of pronounced recent increases in gyrA mutation proxies at a small subset of facilities (Facilities 117, 126, 26, 69, and 30) is epidemiologically significant. Fluoroquinolone resistance represents a critical inflection point in the drug-resistance continuum because fluoroquinolones are core Group A drugs in second-line regimens, and resistance to any fluoroquinolone defines pre-XDR-TB under WHO definitions [ 29 ]. In recent programmatic studies, fluoroquinolone resistance has remained associated with poorer treatment outcomes compared with MDR/RR-TB without FQ resistance, although the magnitude of this effect may vary with access to effective all-oral regimens and timely regimen optimisation [ 30 , 31 ]. The fact that these increases are concentrated within a limited number of facilities suggests localized amplification or selective pressure, rather than widespread community-level dissemination. The facility-level clustering and recent increases are consistent with localized amplification or selective pressure linked to treatment and prescribing pathways; however, recent WGS studies indicate that second-line resistance can also be driven by transmission within connected referral and community networks, which may present as geographically or facility-associated clustering. Further linkage of mutation-proxy trends to treatment history (baseline vs follow-up) and genomic clustering would help distinguish amplification from transmission in this setting [ 32 , 33 ]. Such patterns are unlikely to be detected through conventional aggregate reporting, which typically smooths short-term fluctuations and dilutes facility-specific signals. Similarly, the identification of recent increases in injectable-associated mutation proxies (rrs and/or eis) at a distinct but overlapping set of facilities (Facilities 9, 117, 21, 138, and 11) underscores the capacity of facility-level monitoring to detect early shifts in second-line resistance risk. This finding is consistent with evidence from molecular epidemiology that second-line drug resistance is not evenly distributed but instead may cluster spatially or programmatically where selective pressures or transmission networks are concentrated [ 15 ]. This supports the idea that facility or district characteristics can shape where resistant variants first emerge and are most detectable. Such heterogeneity underscores the potential value of facility-level mutation monitoring as an early-warning mechanism that can detect emerging resistance trends before they become more common at larger geographic scales. However, it is also important to recognise that clustering of resistance does not exclusively indicate de novo amplification; in some high-burden settings, genomic analyses have shown that transmission of resistant strains contributes significantly to observed resistance patterns, which may still appear geographically localized depending on connected transmission networks across facilities and communities [ 28 ]. Distinguishing between amplification and transmission pathways may require linking proxy mutation data to patient treatment histories and performing fine-scale genomic clustering. Although injectable resistance remains relatively uncommon overall, its emergence in specific facilities may reflect treatment-history effects, referral of complex cases, or programmatic challenges such as delayed regimen modification or suboptimal adherence support. Recent laboratory and programmatic studies show that SLID resistance is typically detected at low frequency when broader DR-TB samples are analysed, but clusters appear in higher-risk subgroups. Similarly, multicountry evaluations of Xpert MTB/XDR in sub-Saharan Africa demonstrate that detection of aminoglycoside/injectable resistance targets (rrs/eis-related) occurs in a minority of tested DR-TB cases, reflecting the relatively lower prevalence of these mutations compared with INH- or FQ-associated loci in many routine cohorts [ 25 , 34 ]. Importantly, Facility 117 appeared in both fluoroquinolone- and injectable-associated early-warning analyses, highlighting how convergent resistance signals across drug classes can flag facilities at particularly high risk of progression toward extensively drug-resistant TB. The use of quarterly time-series plots with facility-specific baselines provides a critical interpretive advantage. By visualizing recent changes alongside historical trends, this approach distinguishes meaningful deviations from expected background variability and avoids over-interpretation of single-period spikes. The dashed demarcation of the two most recent quarters facilitates rapid appraisal by programme managers, enabling translation of complex molecular data into an intuitive governance signal. This design aligns with the concept of signal-based surveillance , where the timing and direction of change are as informative as absolute prevalence [ 35 , 36 ]. From a surveillance perspective, these findings support the argument that facility-level mutation surveillance can function as a leading indicator of resistance emergence, potentially preceding formal increases in categorical resistance interpretations [ 37 , 38 ]. Recent genomic surveillance studies similarly emphasize early detection of resistance mutations to guide timely programmatic action, and implementation studies of second-line reflex testing demonstrate the feasibility of embedding such early-warning approaches in routine laboratory workflows [ 39 ]. Because mutation proxies capture underlying genetic changes directly, they may provide earlier insight into evolving resistance dynamics than phenotype-based summaries, which often lag due to diagnostic, reporting, and aggregation delays. In high-burden settings with constrained access to routine sequencing, this approach offers a pragmatic alternative for strengthening resistance intelligence using existing diagnostic infrastructure. Policy and programmatic implications The facility-level mutation fingerprinting framework presented in this study is directly relevant to national and subnational TB control programs. By leveraging routinely generated laboratory molecular diagnostic outputs, this approach enables near–real-time identification of facilities exhibiting emerging drug-resistance signals, without additional laboratory testing or resource investment. Integration of mutation fingerprint surveillance into existing TB monitoring dashboards could support targeted programmatic actions, including intensified adherence support, regimen review, enhanced infection prevention and control, and focused clinical governance at facilities flagged through early-warning indicators. At a policy level, this framework aligns with precision public health principles by shifting drug-resistance surveillance from passive reporting to proactive, data-driven intervention, and could complement existing phenotypic and sequencing-based surveillance systems. Adoption of this approach has the potential to strengthen TB drug-resistance containment strategies, optimize resource allocation, and improve treatment outcomes in high-burden settings. Conclusion This study demonstrates that routine molecular diagnostic data can be repurposed beyond individual-level resistance reporting to support facility-level genomic surveillance of drug-resistant Mycobacterium tuberculosis . By operationalizing mutant melt peak temperatures as mutation-proxy signals and aggregating these data across time and facilities, we identified distinct facility mutation fingerprints, strong associations between mutation proxies and phenotypic resistance, and early-warning signals indicative of emerging resistance risk. The findings highlight substantial heterogeneity in mutational patterns across facilities, suggesting that resistance emergence is shaped by local selective pressures rather than occurring uniformly across the health system. The strong, near-deterministic relationships observed between key mutation proxies and resistance interpretations underscore the robustness of routine molecular outputs as indicators of underlying genetic resistance mechanisms. Importantly, the identification of recent facility-level increases in fluoroquinolone- and injectable-associated mutation proxies illustrates the potential of this approach to detect incipient resistance trends earlier than conventional summary reporting. From a public health perspective, facility-level mutation fingerprinting offers a practical, scalable, and cost-neutral surveillance framework that leverages existing laboratory infrastructure without requiring additional testing or sequencing capacity. Integrating such analyses into routine TB monitoring systems could enable more timely, targeted programmatic responses, including focused adherence support, treatment review, and infection control interventions at facilities exhibiting early warning signals. Future work should explore integrating mutation fingerprint surveillance with patient-level treatment histories, referral pathways, and programmatic interventions, and assess its applicability in other high-burden settings. Overall, this study provides proof of concept for transforming routine molecular diagnostic data into actionable intelligence to strengthen TB drug-resistance surveillance and control. Abbreviations AI: Artificial intelligence AMK: Amikacin CAP: Capreomycin CG: Clinical Governance CI: Confidence interval CXR: Chest X-ray DR-TB: Drug-resistant tuberculosis eis: Enhanced intracellular survival gene ETH: Ethionamide FQ / FLQ: Fluoroquinolones HREC: Health Research Ethics Committee INH: Isoniazid KAN: Kanamycin katG: Catalase-peroxidase gene associated with isoniazid resistance LPA: Line probe assay MDR-TB: Multidrug-resistant tuberculosis MTB: Mycobacterium tuberculosis NHLS: National Health Laboratory Service OR: Odds ratio PCR: Polymerase chain reaction PTM: Pretomanid rpoB: RNA polymerase β-subunit gene rrs: 16S ribosomal RNA gene associated with aminoglycoside resistance SLID: Second-line injectable drug TB: Tuberculosis WGS: Whole-genome sequencing WSU: Walter Sisulu University XDR-TB: Extensively drug-resistant tuberculosis Xpert MTB/XDR: Molecular diagnostic assay for resistance detection Declarations Ethics approval and consent to participate This study complied with the ethical principles of the Declaration of Helsinki. Ethical approval was obtained from the Research Ethics Biosafety Committee, Faculty of Health Sciences, Walter Sisulu University (WSU HREC 141/2025; approved July 2, 2025). The Eastern Cape Department of Health granted administrative authorization to access and analyse facility-level data (Ref: EC_202507_023; approved July 11, 2025). The National Health Laboratory Service approved the use of de-identified laboratory diagnostic data for research purposes (SR4169693; approved November 17, 2025). Because the study used de-identified secondary data generated during routine diagnostic services, with no direct patient contact or additional specimen collection, the ethics committee approved the study and waived the requirement for individual informed consent. Consent for publication Not applicable. This study uses de-identified secondary data with no individual-level identifiable information. Clinical trial registration Clinical trial number: Not applicable. This study did not involve an interventional clinical trial and, therefore, did not require trial registration. Availability of data and materials The datasets generated and analyzed during the current study are not publicly available due to data protection regulations of the Eastern Cape Department of Health and the National Health Laboratory Service. De-identified aggregated data may be made available from the corresponding author upon reasonable request and with permission of the respective data custodians. Competing interests The authors are pleased to share that they have no competing interests to disclose that relate to the content of this article. Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Institutional support was provided by Walter Sisulu University, Faculty of Health Sciences. Author contributions Lindiwe Modest Faye : Conceptualization, methodology, supervision, formal analysis, data interpretation, original draft writing, review, and editing. Melisa Makhuba: Investigation, data curation, laboratory work, resources, formal analysis, review, and editing. Ntandazo Dlatu : Software, statistical analysis, validation, visualization, review, and editing. Mojisola Clara Hosu: Literature review, methodology, project administration, data interpretation, review, and editing. Teke Apalata : Supervision, clinical and laboratory oversight, funding acquisition, project administration, review and editing, and final approval of the manuscript. Acknowledgements The authors acknowledge the Eastern Cape Department of Health, the National Health Laboratory Service, and the participating health facilities for their cooperation. We also thank the Walter Sisulu University TB Research Unit for technical and administrative support. References Auld SC, Shah NS, Mathema B, Brown TS, Ismail N, Omar SV, Brust JCM, Nelson KN, Allana S, Campbell A, Mlisana K. Extensively drug-resistant tuberculosis in South Africa: genomic evidence supporting transmission in communities. Eur Respir J. 2018;52(4):1800246. https://doi.org/10.1183/13993003.00246-2018 . Naidoo K, Perumal R, Cox H, Mathema B, Loveday M, Ismail N, Omar SV, Georghiou SB, Daftary A, O'Donnell M, Ndjeka N. The epidemiology, transmission, diagnosis, and management of drug-resistant tuberculosis: lessons from the South African experience. Lancet Infect Dis. 2024;24(9):e559–75. https://doi.org/10.1016/S1473-3099(24)00144-0 . Elbehiry A, Marzouk E, Edrees HM, AlShaqi R, Ellethy AT, Alzaben F, Anagreyyah S, Algarni A, Almuhaydili K, Alotaibi I, Albaqami A. Advancing the fight against tuberculosis: integrating innovation and public health in diagnosis, treatment, vaccine development, and implementation science. Front Med. 2025;12:1596579. https://doi.org/10.3389/fmed.2025.1596579 . Ullah MN, Khan HR, Imran A, Saeed A, Yasin U, Ahmed N, Shah SAH. The future of Mycobacterium tuberculosis diagnostics: integrating mass spectrometry, AI, and genomics. J Health Wellness Community Res. 2025;e334. https://doi.org/10.61919/0fvx1h30 . Cao Y, Parmar H, Simmons AM, Kale D, Tong K, Lieu D, Persing D, Kwiatkowski R, Alland D, Chakravorty S. Automatic identification of rpoB mutations responsible for rifampin resistance using melting temperature signatures with Xpert MTB/RIF Ultra. J Clin Microbiol. 2019;58(1):10–1128. https://doi.org/10.1128/jcm.00907-19 . Torrea G, Ng KC, Van Deun A, André E, Kaisergruber J, Ssengooba W, Desmaretz C, Gabriels S, Driesen M, Diels M, Asnong S. Variable ability of rapid tests to detect Mycobacterium tuberculosis rpoB mutations conferring phenotypically occult rifampicin resistance. Sci Rep. 2019;9:11826. https://doi.org/10.1038/s41598-019-48401-z . Garcia LS, Costa AG, Araújo-Pereira M, Spener-Gomes R, França Aguiar A, Souza AB, Lima LO, Benjamin A, Rocha MS, Moreira AS, Silva J. Xpert MTB/RIF cycle threshold predicts TB transmission to close contacts in Brazil. Clin Infect Dis. 2024;ciad794. Guo Y, Yang J, Wang H, Sha W, Yu F. Key resistance-associated mutations in MDR-TB with focus on aminoglycosides: a genomic study from Shanghai, China. BMC Microbiol. 2025;25:702. https://doi.org/10.1186/s12866-025-04446-x . Singh AK, Singh N, Kumar S, Mishra AK, Singh NP. Molecular insights of drug-resistant tuberculosis: genetic mutations and their profile. Front Microbiol. 2025;16:1669327. https://doi.org/10.3389/fmicb.2025.1669327 . Gilmour B, Alene KA. Ending tuberculosis: challenges and opportunities. Front Tuberc. 2024;2:1487518. https://doi.org/10.3389/ftubr.2024.1487518 . Ding F, Liu W, Wu C, Zhang W, Chen S, Lai W, Qu J, Lin Q, Lu S, Qu J. Whole-genome sequencing reveals TB transmission patterns and drug resistance within/between hosts. Front Cell Infect Microbiol. 2025;14:1488547. https://doi.org/10.3389/fcimb.2024.1488547 . Gopalaswamy R, Subbian S. The power of resistance: mechanisms of antimicrobial resistance in Mycobacterium tuberculosis and impact on TB management. Clin Microbiol Rev. 2026. https://doi.org/10.1128/cmr.00194-25 . e00194-25. de Zwaan R, de Vries G, Ubbelohde E, Mulder A, Kamst-van Agterveld M, Rebel K, Kautz S, Kremer K, Anthony RM, van Soolingen D. Verification of emerging genomic mutations enables distinguishing TB transmission chains over 30 years. PLoS ONE. 2025;20(6):e0319630. https://doi.org/10.1371/journal.pone.0319630 . Marcon DJ, Sharma A, Souza AB, Barros RB, Andrade VDGD, Guimarães RJDPS, Lima LNG, Monteiro LHMT, Quaresma AJPG, Ribeiro LR, Suffys PN. Genomic surveillance reveals XDR-TB transmission profiles in Pará, Brazil. Front Microbiol. 2025;15:1514862. https://doi.org/10.3389/fmicb.2024.1514862 . Ngom JT, Loubser J, Maasdorp E, Ghebrekristos Y, Singh S, Opperman CJ, Klopper M, Warren RM, Streicher EM. Population structure & emerging resistance to new TB drugs in Western Cape: 10-year genomic study. Front Cell Infect Microbiol. 2025;15:1638577. https://doi.org/10.3389/fcimb.2025.1638577 . Nono VN, Nantia EA, Mutshembele A, Teagho SN, Simo YWK, Takong BS, Djieugoue YJ, Assolo YP, Ongboulal SM, Awungafac SN, Eyangoh S. Prevalence of katG and inhA mutations linked to isoniazid resistance in Cameroon. BMC Microbiol. 2025;25(1):127. https://doi.org/10.1186/s12866-025-03816-9 . Pei S, Song Z, Yang W, He W, Ou X, Zhao B, He P, Zhou Y, Xia H, Wang S, Jia Z. Catalogue of M. tuberculosis mutations associated with drug resistance to 12 drugs in China: nationwide genomic study. Lancet Microbe. 2024;5(11):100899. https://doi.org/10.1016/S2666-5247(24)00131-9 . Sy KTL, Leavitt SV, de Vos M, Dolby T, Bor J, Horsburgh CR Jr, Warren RM, Streicher EM, Jenkins HE, Jacobson KR. Spatial heterogeneity of XDR-TB in Western Cape Province, South Africa. Sci Rep. 2022;12:10844. https://doi.org/10.1038/s41598-022-14581-4 . Diriba G, Mollalign H, Meaza A, Getu M, Zerihun B, Alemu A, Dagne B, Abebaw Y, Weldemariam AG, Bashea C, Ali A. WGS-based detection of XDR-TB in Ethiopia. Commun Med. 2026;6:14. https://doi.org/10.1038/s43856-025-01271-1 . Valafar SJ. Systematic review of mutations associated with isoniazid resistance points to continuing evolution and subsequent evasion of molecular detection, and potential for emergence of multidrug resistance in clinical strains of Mycobacterium tuberculosis. Antimicrob Agents Chemother. 2021;65(3):10–128. Ngema SL, Dookie N, Perumal R, Nandlal L, Naicker N, Letsoalo MP, O'Donnell M, Khan A, Padayatchi N, Naidoo K. Isoniazid resistance-conferring mutations are associated with highly variable phenotypic resistance. J Clin Tuberc Other Mycobact Dis. 2023;33:100387. Dean AS, Auguet OT, Glaziou P, Zignol M, Ismail N, Kasaeva T, Floyd K. 25 years of surveillance of drug-resistant tuberculosis: achievements, challenges, and way forward. Lancet Infect Dis. 2022;22(7):e191–6. Liebenberg D, Gordhan BG, Kana BD. Drug-resistant tuberculosis: implications for transmission, diagnosis, and disease management. Front Cell Infect Microbiol. 2022;12:943545. Xiao YX, Liu KH, Lin WH, Chan TH, Jou R. Whole-genome sequencing-based analyses of drug-resistant Mycobacterium tuberculosis from Taiwan. Sci Rep. 2023;13:2540. https://doi.org/10.1038/s41598-023-29652-3 . Oliveira RS, Brandao AP, Ferreira FMDA, Costa SMD, Silva VLM, Ferrazoli L, Chimara E, Pinhata JMW. Fluoroquinolone and second-line injectable resistance among RIF- and INH-resistant M. tuberculosis. Microorganisms. 2025;13:2470. https://doi.org/10.3390/microorganisms13112470 . Timm J, Bateson A, Solanki P, Paleckyte A, Witney AA, Rofael SA, Fabiane S, Olugbosi M, McHugh TD, Sun E. Baseline and acquired resistance to BDQ, LZD and PTM in four clinical trials. PLOS Glob Public Health. 2023;3:e0002283. https://doi.org/10.1371/journal.pgph.0002283 . Mboowa G. Reimagining TB control in the genomics era: global investment in TB genomic surveillance. Pathogens. 2025;14:975. https://doi.org/10.3390/pathogens14100975 . Sobol R, Omar SV, Brown T, Joseph L, Lutchminarian K, Tang L, Lan Y, Willis F, Campbell A, Warren JL, Cohen T. Transmission of BDQ-resistant TB in KwaZulu-Natal. Am J Respir Crit Care Med. 2025. https://doi.org/10.1164/rccm.202506-1489OC . World Health Organization. WHO announces updated definitions of extensively drug-resistant tuberculosis. 2021. Available from: https://www.who.int/news/item/27-01-2021-who-announces-updated-definitions-of-extensively-drug-resistant-tuberculosis Nehru VJ, Jose Vandakunnel M, Brammacharry U, Ramachandra V, Pradhabane G, Mani BR, Vn AD, Muthaiah M. FQ resistance transmission in DR-TB: genomic epidemiology study. Sci Rep. 2024;14:19719. https://doi.org/10.1038/s41598-024-70535-y . Kim S, Mok J. Treatment outcomes of FQ-resistant MDR-TB: implications for delamanid—authors’ reply. Tuberc Respir Dis. 2024;87:209–11. https://doi.org/10.4046/trd.2024.0010 . Li M, Zhang Y, Wu Z, Jiang Y, Sun R, Yang J, Li J, Lin H, Zhang R, Jiang Q, Wang L. Transmission of FQ resistance in MDR-TB in Shanghai: WGS epidemiology. Emerg Microbes Infect. 2024;13:2302837. https://doi.org/10.1080/22221751.2024.2302837 . Zhang Y, Li D, Liu J, Jiang Y, Shen X, Xu B. Pre-diagnostic FQ exposure & acquired resistance in M. tuberculosis: WGS-based case-control study. J Infect. 2025;91:106648. Massou F, Diarra B, Diallo AB, Bah KS, Vuchas C, Neh A, Sanders M, Adebiyi EO, Aderemi BO, Agbla SC, Flyod S. Multicenter evaluation of Xpert MTB/XDR in Sub-Saharan Africa. ERJ Open Res. 2025. https://doi.org/10.1183/23120541.00427-2025 . Rilkoff H, Struck S, Ziegler C, Faye L, Paquette D, Buckeridge D. Innovations in public health surveillance: novel data & analytic methods. Can Commun Dis Rep. 2024;50:93–101. https://doi.org/10.14745/ccdr.v50i34a02 . Hicketier A, Bach M, Oedi P, Ullrich A, Abbood A. Ensemble-labeling of infectious disease time series to evaluate early warning systems. Infect Dis Model. 2025. Barilar I, Fernando T, Utpatel C, Abujate C, Madeira CM, José B, Mutaquiha C, Kranzer K, Niemann T, Ismael N, de Araujo L. Emergence of BDQ-resistant TB and undetected rpoB Ile491Phe MDR/XDR-TB in Mozambique. Lancet Infect Dis. 2024;24:297–307. https://doi.org/10.1016/S1473-3099(23)00498-X . Makondo VT, Kaapu KG, Wells F, Sharma A, Lekalakala-Mokaba MR, Warren R, Conceição EC, Rukasha I. Empowering TB genomic surveillance in Limpopo, South Africa. Front Public Health. 2025;13:1567382. https://doi.org/10.3389/fpubh.2025.1567382 . Centner CM, Munir R, Tagliani E, Rieß F, Brown P, Hayes C, Dolby T, Zemanay W, Cirillo DM, David A, Schumacher SG. Reflex Xpert MTB/XDR testing of RR-TB specimens: diagnostic accuracy and feasibility. Open Forum Infect Dis. 2024;11:ofae437. https://doi.org/10.1093/ofid/ofae437 . Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 26 Feb, 2026 Reviewers invited by journal 20 Feb, 2026 Editor invited by journal 10 Feb, 2026 Editor assigned by journal 08 Feb, 2026 Submission checks completed at journal 08 Feb, 2026 First submitted to journal 04 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8790685","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":594640145,"identity":"fdb5d33c-13d2-4161-baae-abdbf1be6548","order_by":0,"name":"Lindiwe Modest Faye","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAvElEQVRIiWNgGAWjYBACPuYDQLKCFC1sbAlA8gzJWhjbSNPC/Ezi47zD9gYHmB9+YPhjR4wWNjPJmdsOJ244wGYswcCTTIQW+QYzad5ttxMMDjCYMTBIMBNjC/s36b9zbgMdxv6NgcGgnhgtPGbSjA23GTcc4AHaknCYKC3Flj3H/ifOPMxTLJFw4DhhLfxs7Btv/KhJs+c73r7xw4c/1YS1AAGLBJgCeTyBKA1AtR+IVDgKRsEoGAUjFQAAZNMxPPWgI4QAAAAASUVORK5CYII=","orcid":"","institution":"Walter Sisulu University","correspondingAuthor":true,"prefix":"","firstName":"Lindiwe","middleName":"Modest","lastName":"Faye","suffix":""},{"id":594640148,"identity":"ec57d2b5-1279-4d9a-b06b-d61ce085f094","order_by":1,"name":"Melisa Makhuba","email":"","orcid":"","institution":"Walter Sisulu University","correspondingAuthor":false,"prefix":"","firstName":"Melisa","middleName":"","lastName":"Makhuba","suffix":""},{"id":594640151,"identity":"ba1e1d9e-479b-489e-9619-026ef4c90247","order_by":2,"name":"Ntandazo Dlatu","email":"","orcid":"","institution":"Healthcare Administration, Walter Sisulu University","correspondingAuthor":false,"prefix":"","firstName":"Ntandazo","middleName":"","lastName":"Dlatu","suffix":""},{"id":594640152,"identity":"688d2ebe-ef36-4f98-8677-e4d6110d8cfb","order_by":3,"name":"Mojisola Clara Hosu","email":"","orcid":"","institution":"Walter Sisulu University","correspondingAuthor":false,"prefix":"","firstName":"Mojisola","middleName":"Clara","lastName":"Hosu","suffix":""},{"id":594640153,"identity":"aeea3f0c-883c-414f-821e-042833ca4748","order_by":4,"name":"Teke Apalata","email":"","orcid":"","institution":"Walter Sisulu University","correspondingAuthor":false,"prefix":"","firstName":"Teke","middleName":"","lastName":"Apalata","suffix":""}],"badges":[],"createdAt":"2026-02-04 23:08:54","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8790685/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8790685/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103445097,"identity":"3b2dce28-7fbc-4767-8aef-52b624cae893","added_by":"auto","created_at":"2026-02-25 18:29:32","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":66278,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of drug-resistance classified as susceptible (S), resistant (R), low-level resistant (LR), uninterpretable (US/IUS), or not tested (NT) in percentages for INH, FLQ, AMK, KAN, CAP, and ETH.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8790685/v1/14987ec1e7fdd4dec1f20cae.png"},{"id":103508211,"identity":"a0fa088c-e858-4ffe-8182-0708d3277aa4","added_by":"auto","created_at":"2026-02-26 13:47:35","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":76107,"visible":true,"origin":"","legend":"\u003cp\u003eFacility-level mutation fingerprint heatmap displaying prevalence\u003cem\u003e (%) \u003c/em\u003eof mutation-proxy signals\u003cem\u003e (katG, inhA, gyrA-any, rrs, eis-\u003c/em\u003eany, injectable-any\u003cem\u003e) \u003c/em\u003eacross facilities, clustered by similarity of mutation profiles.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8790685/v1/8d15b92036479614020afbc1.png"},{"id":103445098,"identity":"adf828ba-22df-4707-9ae3-895dfdbb884c","added_by":"auto","created_at":"2026-02-25 18:29:32","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":133570,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEarly-warning time-series of \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003egyrA\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e mutation-proxy prevalence by facility\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8790685/v1/6bbc122f8a43d5e544ca555f.png"},{"id":103445100,"identity":"6acf404c-92d7-4a88-aaa6-e6e6f142df4b","added_by":"auto","created_at":"2026-02-25 18:29:32","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":129746,"visible":true,"origin":"","legend":"\u003cp\u003eEarly-warning time-series of injectable-associated mutation-proxy prevalence by facility\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8790685/v1/83355752efe3324c0c004283.png"},{"id":103511778,"identity":"1857aca7-0a3a-4cb1-8cb4-cb673a3f2a2f","added_by":"auto","created_at":"2026-02-26 14:10:58","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1595934,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8790685/v1/b4e77ec0-79b0-4b5f-b037-7ca2cbf1ec51.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Facility-Level Mutation Fingerprints and Early-Warning Surveillance of Drug-Resistant Mycobacterium tuberculosis in Rural Eastern Cape","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eTuberculosis (TB) remains a formidable public health challenge in South Africa, exacerbated by the persistent rise of drug resistance and unabated transmission within both healthcare and community settings [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. The advent of molecular diagnostic platforms has revolutionized TB detection, enabling rapid identification of resistance-associated mutations to guide clinical decision-making [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. However, a significant gap remains: the high-resolution molecular outputs generated by these platforms are primarily restricted to individual patient management, leaving their potential for population-level surveillance and health-system monitoring largely untapped.\u003c/p\u003e \u003cp\u003eRoutine molecular assays employed in the diagnosis of \u003cem\u003eMycobacterium tuberculosis\u003c/em\u003e produce more than just binary results; they generate rich, high-resolution proxy signals of genetic variation, including mutant and wild-type melt peak temperatures and cycle threshold values [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. These signals directly reflect underlying mutations in critical resistance-associated loci, such as \u003cem\u003ekatG\u003c/em\u003e, \u003cem\u003einhA\u003c/em\u003e, \u003cem\u003egyrA\u003c/em\u003e, \u003cem\u003errs\u003c/em\u003e, and \u003cem\u003eeis\u003c/em\u003e [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Despite being generated at scale, these data are rarely synthesized to examine the spatial, temporal, or facility-level patterns of mutational burden and the kinetics of resistance emergence. Conventional TB drug-resistance surveillance currently relies on aggregated phenotypic outcomes or categorical molecular interpretations (e.g., \"Susceptible\" vs. \"Resistant\").\u003c/p\u003e \u003cp\u003eWhile these metrics are essential, they are inherently reactive and may obscure the subtle, early shifts in mutational landscapes that precede widespread clinical resistance. In high-burden settings, healthcare facilities and districts often exhibit distinct patient case-mixes, treatment histories, and referral patterns. These unique local dynamics impose heterogeneous selective drug pressures, driving facility-specific trajectories of resistance emergence and circulation that are invisible to standard reporting mechanisms [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eRepurposing these routine molecular outputs for facility-level analysis represents a paradigm shift from passive reporting to proactive genomic proxy surveillance. By aggregating these signals over time and geography, we can define \"mutation fingerprints\" that characterize the dominant and emerging resistance profiles of specific healthcare sites. These fingerprints provide a critical window for intervention, serving as early-warning indicators that allow for targeted programmatic responses before resistance becomes entrenched within a facility or community [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis study, therefore, aimed to develop and evaluate a facility-level mutation fingerprinting and early-warning surveillance framework using routine molecular diagnostic data. By integrating mutation-proxy signals, resistance interpretations, and temporal trends with health-system metadata, we demonstrate how underutilized laboratory data can be transformed into actionable intelligence to strengthen TB control.\u003c/p\u003e"},{"header":"2. Methodology","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Study design\u003c/h2\u003e \u003cp\u003eA retrospective analytical study was conducted using routinely collected molecular diagnostic data for Mycobacterium tuberculosis. The study applied a mixed descriptive, inferential, and exploratory modelling approach to characterize facility-level mutation patterns and assess their relationship with interpretations of drug resistance over time.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Data source and study population\u003c/h2\u003e \u003cp\u003eThe dataset comprised TB diagnostic records generated through laboratory molecular diagnostic testing platforms. Each record corresponded to a patient diagnostic sample and included unique patient and identifiers, specimen collection dates, facility and geographic metadata, molecular assay outputs, and drug-resistance interpretation results. All records with valid specimen dates and facility identifiers were eligible for inclusion. Records with incomplete key identifiers were excluded from facility-level aggregation but retained for descriptive summaries where appropriate.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Variables and operational definitions\u003c/h2\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003e2.3.1 Facility and geographic variables\u003c/h2\u003e \u003cp\u003eFacility-level analysis was conducted using routinely recorded geographic and service-related variables. These variables were used to characterize the spatial distribution of diagnostic episodes and to support aggregation of molecular surveillance data across health system levels. For this study, health-care facilities were treated as the primary unit of surveillance analysis, enabling facility-level aggregation of mutation patterns, temporal trends, and early-warning signals.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.3.2 Temporal variables\u003c/h2\u003e \u003cp\u003eSpecimen collection dates were used to derive monthly and quarterly time periods for temporal trend and early-warning analyses.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003e2.3.3 Mutation-proxy variables\u003c/h2\u003e \u003cp\u003eMutation-proxy signals were derived from mutant melt peak temperature outputs for resistance-associated loci, including \u003cem\u003ekatG\u003c/em\u003e and \u003cem\u003einhA\u003c/em\u003e (isoniazid-associated), \u003cem\u003egyrA\u003c/em\u003e regions (fluoroquinolone-associated), and \u003cem\u003errs\u003c/em\u003e and \u003cem\u003eeis\u003c/em\u003e (injectable-associated). For each locus, mutation presence was operationalized as a binary indicator, with mutant melt peak temperatures coded as 1 (detected) and absence coded as 0. Wild-type melt peak temperatures and cycle threshold values were retained as continuous variables to support the assessment of signal characteristics.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.3.4 Drug-resistance interpretation variables\u003c/h2\u003e \u003cp\u003eResistance interpretation variables were extracted for isoniazid (INH), fluoroquinolones (FLQ), amikacin (AMK), kanamycin (KAN), capreomycin (CAP), and ethionamide (ETH). Standard interpretation categories, namely susceptible, resistant, low-level resistant, uninterpretable, and not tested, were analyzed descriptively and incorporated as outcome variables in regression modelling.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Construction of facility mutation fingerprints\u003c/h2\u003e \u003cp\u003eFacility mutation fingerprints were constructed by aggregating mutation-proxy indicators at the facility\u0026ndash;time level. For each facility and time, mutation prevalence was calculated as the proportion of tested episodes with a detected mutant melt peak. Facility fingerprints were represented as vectors of mutation prevalence across loci, enabling comparisons between facilities and overtime.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Exploratory clustering of facilities\u003c/h2\u003e \u003cp\u003eUnsupervised clustering techniques were applied to facility mutation fingerprints to identify groups of facilities with similar mutational profiles. Hierarchical clustering using appropriate distance measures for proportional data was employed. Cluster robustness was evaluated using internal validation metrics, and clusters were interpreted in relation to geographic location and resistance interpretation patterns.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e2.6 Early-warning and temporal trend analysis\u003c/h2\u003e \u003cp\u003eTo detect emerging resistance signals, time-series analyses were conducted on key mutation proxies at the facility level. Baseline prevalence estimates were established for each facility, and deviations from baseline were assessed using statistical change-detection approaches. Facilities exhibiting sustained or abrupt increases in high-risk mutation proxies were flagged as potential early-warning signals. These signals were examined alongside resistance interpretation outcomes to assess concordance and lead-time advantages over conventional reporting.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e2.7 Association between mutation fingerprints and resistance outcomes\u003c/h2\u003e \u003cp\u003eMixed-effects regression models were fitted to quantify associations between mutation-proxy patterns and drug-resistance interpretation outcomes. Facility was included as a random effect to account for clustering of episodes within facilities. Fixed effects included mutation-proxy indicators and selected temporal covariates. Model outputs were used to assess the strength and consistency of relationships between facility-level mutational burden and observed resistance patterns.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e2.8 Statistical analysis\u003c/h2\u003e \u003cp\u003eAll analyses were conducted using reproducible, script-based workflows. Data cleaning and preprocessing included verification of unique patient and diagnostic episode identifiers, harmonization of facility identifiers, and transformation of specimen collection dates into standardized monthly and quarterly time periods.\u003c/p\u003e \u003cp\u003eMutation-proxy variables were operationalized as binary indicators based on the presence or absence of mutant melt peak temperatures for resistance-associated loci (\u003cem\u003ekatG\u003c/em\u003e, \u003cem\u003einhA\u003c/em\u003e, \u003cem\u003egyrA\u003c/em\u003e regions, \u003cem\u003errs\u003c/em\u003e, and \u003cem\u003eeis\u003c/em\u003e). Wild-type melt peak temperatures and cycle threshold (CT) values were retained as continuous variables for descriptive assessment and sensitivity analyses.\u003c/p\u003e \u003cp\u003eFacility-level mutation fingerprints were generated by aggregating mutation-proxy prevalence within facilities across defined time periods. Unsupervised clustering of facilities was performed using hierarchical clustering with Ward\u0026rsquo;s linkage on standardized mutation-prevalence vectors. Cluster structure and robustness were assessed using internal validation metrics and visual inspection of dendrograms.\u003c/p\u003e \u003cp\u003eTemporal trends and early-warning signals were evaluated by analyzing facility-level mutation prevalence over time. Baseline prevalence estimates were established for each facility, and deviations from baseline were identified using change-detection approaches appropriate for proportion-based time-series data.\u003c/p\u003e \u003cp\u003eAssociations between mutation-proxy signals and drug-resistance interpretation outcomes were examined using mixed-effects logistic regression models. Resistance interpretation outcomes were treated as categorical dependent variables, with mutation-proxy indicators specified as fixed effects. Facility was included as a random intercept to account for clustering of diagnostic episodes within facilities. Model results were reported as adjusted odds ratios with corresponding 95% confidence intervals.\u003c/p\u003e \u003cp\u003eAll statistical analyses were conducted using R (version 4.3). Key packages included \u003cem\u003etidyverse\u003c/em\u003e for data management and visualization, \u003cem\u003elme4\u003c/em\u003e for mixed-effects modelling, \u003cem\u003ecluster\u003c/em\u003e and \u003cem\u003efactoextra\u003c/em\u003e for clustering analyses, and \u003cem\u003echangepoint\u003c/em\u003e for temporal change detection. Statistical significance was assessed using a two-sided alpha level of 0.05.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e2.9 Sample size and power considerations\u003c/h2\u003e \u003cp\u003eThis study used secondary routine laboratory data comprising approximately 4,300 \u003cem\u003eMycobacterium tuberculosis\u003c/em\u003e diagnostic patient samples aggregated across 139 health-care facilities. As the dataset represents near-complete routine testing during the study period rather than a sampled population, a formal a priori sample size calculation was not applicable.\u003c/p\u003e \u003cp\u003eNevertheless, post hoc power and precision considerations indicate that the available sample size was adequate for the planned analyses. With over 4,000 patients, the study had sufficient statistical power (\u0026gt;\u0026thinsp;80%) to detect small-to-moderate associations (odds ratios of approximately 1.3\u0026ndash;1.5 or greater) between mutation-proxy indicators and drug-resistance interpretation outcomes in logistic regression models at a two-sided α level of 0.05.\u003c/p\u003e \u003cp\u003eAt the facility level, the inclusion of 139 facilities provided adequate variability to support facility-level aggregation, clustering, and mixed-effects modelling. The number of facilities exceeded recommended thresholds for stable estimation of random effects in multilevel models, allowing reliable assessment of between-facility heterogeneity in mutation patterns and resistance outcomes.\u003c/p\u003e \u003cp\u003eFor early-warning and temporal trend analyses, the large number of observations per facility over time enabled detection of meaningful deviations in mutation-proxy prevalence beyond expected random fluctuation. Facilities with low testing volumes were handled by aggregating across quarterly intervals and performing sensitivity analyses to ensure the robustness of detected signals.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Study dataset and facility coverage\u003c/h2\u003e \u003cp\u003eA total of 4,300 tuberculosis diagnostic patients\u0026rsquo; samples were included in the analysis, representing routine molecular testing conducted across 139 health-care facilities. All records contained valid facility identifiers and specimen collection dates, allowing for facility-level aggregation and temporal analyses. The dataset spanned 16 calendar quarters, enabling assessment of longitudinal trends and recent changes in mutation prevalence.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Prevalence of mutation-proxy signals\u003c/h2\u003e \u003cp\u003eMutation-proxy signals were identified by the presence of mutant melt peak temperatures at resistance-associated loci. Overall, isoniazid-associated mutation proxies predominated, while fluoroquinolone- and injectable-associated proxies were less frequent. Mutant melt peaks in \u003cem\u003ekatG\u003c/em\u003e were detected in 46.3% of diagnostic patients\u0026rsquo; samples, followed by \u003cem\u003emutant melt peaks in inhA at\u003c/em\u003e 25.1%. Mutation proxies in \u003cem\u003egyrA\u003c/em\u003e regions were identified in 12.5% of episodes, and \u003cem\u003errs\u003c/em\u003e-associated mutant peaks were detected in 7.7%. No independent \u003cem\u003eeis\u003c/em\u003e mutant peaks were observed; however, injectable-associated mutation proxies (\u003cem\u003errs\u003c/em\u003e and/or \u003cem\u003eeis\u003c/em\u003e) were present in 7.7% of diagnostic patients\u0026rsquo; samples overall. The prevalence of mutation-proxy signals across all analysed diagnostic patient samples is summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePrevalence of mutation-proxy signals detected by routine molecular testing (N\u0026thinsp;=\u0026thinsp;4,300)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMutation Proxy (Gene/Locus)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNumber of diagnostic patient samples\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePercentage (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003ekatG\u003c/em\u003e mutant melt peak is present\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 991\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e46.3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003einha\u003c/em\u003e mutant melt peak is present\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 080\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e25.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003egyrA\u003c/em\u003e region mutant peak present (any)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e538\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e12.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003errs\u003c/em\u003e mutant melt peak is present\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e331\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e7.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eei\u003c/em\u003es mutant melt peak present\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInjectable-associated proxy (\u003cem\u003errs/eis\u003c/em\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e331\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e7.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Drug-resistance profiles\u003c/h2\u003e \u003cp\u003eThe resistance interpretation in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e demonstrates marked heterogeneity in resistance patterns across first- and second-line anti-tuberculosis drugs. Isoniazid (INH) shows the highest burden of resistance, with 51.6% of diagnostic patients\u0026rsquo; samples classified as resistant and an additional 17.3% as low-level resistant, indicating that nearly 7 in 10 diagnostic patients\u0026rsquo; samples exhibit some degree of isoniazid resistance. Only 21.9% of diagnostic samples from patients were fully susceptible to INH, highlighting the central role of isoniazid resistance in our study.\u003c/p\u003e \u003cp\u003eIn contrast, fluoroquinolones (FLQ) and injectable agents (amikacin, kanamycin, and capreomycin) remained predominantly susceptible, with susceptibility proportions ranging from 76.8% to 83.5%. Resistance to these second-line drugs was comparatively uncommon, occurring in approximately 7\u0026ndash;8% of episodes, and low-level resistance was rare or absent. These patterns suggest that while second-line resistance is present, it has not yet become widespread.\u003c/p\u003e \u003cp\u003eEthionamide (ETH) displayed an intermediate resistance profile, with 25.5% of diagnostic patients\u0026rsquo; samples classified as resistant and 66.1% susceptible, consistent with known cross-resistance between ethionamide and isoniazid mediated through \u003cem\u003einhA\u003c/em\u003e-associated mechanisms.\u003c/p\u003e \u003cp\u003eAcross all drugs, a consistent \u0026ldquo;not tested\u0026rdquo; (NT) category of 8.1% reflects routine diagnostic workflows rather than biological resistance. Uninterpretable and uninterpretable-susceptible categories were infrequent, indicating generally robust assay performance.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Facility-level mutation fingerprints\u003c/h2\u003e \u003cp\u003eFacility-level mutation fingerprints were constructed by aggregating mutation-proxy prevalence across facilities with sufficient testing volume (\u0026ge;\u0026thinsp;30 episodes). Hierarchical clustering revealed marked heterogeneity in mutational profiles between facilities (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Some facilities were characterized by high prevalence of \u003cem\u003ekatG\u003c/em\u003e-dominant mutation patterns, while others showed \u003cem\u003einhA\u003c/em\u003e-dominant or mixed profiles. A smaller subset of facilities demonstrated an elevated prevalence of gyrA-associated mutation proxies, suggesting increased fluoroquinolone resistance risk. Injectable-associated mutation proxies were concentrated in a limited number of facilities.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Association between mutation proxies and drug resistance\u003c/h2\u003e \u003cp\u003eBecause mutation-proxy variables exhibited strong associations with resistance interpretation outcomes, standard logistic and mixed-effects models showed evidence of near-separation, leading to unstable coefficient estimates. To address this, L2-regularised logistic regression was employed, combined with facility-level cluster bootstrap confidence intervals, which provide stable and interpretable estimates under conditions of quasi-complete separation while accounting for clustering of diagnostic episodes within facilities.\u003c/p\u003e \u003cdiv id=\"Sec22\" class=\"Section3\"\u003e \u003ch2\u003e3.5.1 Isoniazid resistance\u003c/h2\u003e \u003cp\u003eFor isoniazid resistance, models including \u003cem\u003ekatG\u003c/em\u003e and \u003cem\u003einhA\u003c/em\u003e mutation proxies demonstrated powerful associations with resistance outcomes. The presence of a \u003cem\u003ekatG\u003c/em\u003e mutant proxy was associated with an adjusted odds ratio (OR) of approximately 1,146.6 (cluster-bootstrap 95% CI: 584.0\u0026ndash;1,928.9), while \u003cem\u003einhA\u003c/em\u003e mutation proxies were associated with an OR of approximately 603.6 (95% CI: 362.3\u0026ndash;885.7) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAssociation between isoniazid resistance and mutation-proxy signals\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMutation proxy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAdjusted OR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95% CI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInterpretation\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003ekatG\u003c/em\u003e mutant proxy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1 146.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e584.0\u0026ndash;1 928.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStrong association with INH resistance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eInhA\u003c/em\u003e mutant proxy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e603.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e362.3\u0026ndash;885.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStrong association with INH resistance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003e3.5.2 Fluoroquinolone resistance\u003c/h2\u003e \u003cp\u003eFor fluoroquinolone resistance, \u003cem\u003egyrA\u003c/em\u003e mutation proxies were the dominant predictors, with an adjusted OR of approximately 7 136.5 (95% CI: 1 348.3\u0026ndash;18 367.0). Secondary associations were observed for \u003cem\u003ekatG\u003c/em\u003e (OR\u0026thinsp;\u0026asymp;\u0026thinsp;4.93, 95% CI: 1.09\u0026ndash;13.78) and \u003cem\u003einhA\u003c/em\u003e (OR\u0026thinsp;\u0026asymp;\u0026thinsp;7.16, 95% CI: 1.49\u0026ndash;21.04), indicating that fluoroquinolone resistance frequently occurred in the context of broader resistance-associated mutation backgrounds (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAssociation between fluoroquinolone resistance and mutation-proxy signals\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMutation proxy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAdjusted OR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95% CI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInterpretation\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003egyrA\u003c/em\u003e mutant proxy (any region)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7 136.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1 348.3\u0026ndash;18 367.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePrimary predictor of FLQ resistance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003ekatG\u003c/em\u003e mutant proxy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.09\u0026ndash;13.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSecondary association\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003einhA\u003c/em\u003e mutant proxy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7.16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.49\u0026ndash;21.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSecondary association\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e3.6 Early-warning signals for emerging resistance\u003c/h2\u003e \u003cp\u003eQuarterly facility-level monitoring identified facilities with the largest recent increases in mutation-proxy prevalence between the two most recent quarters. For fluoroquinolone-associated risk, as indicated by increases in \u003cem\u003egyrA\u003c/em\u003e mutation proxies, the most pronounced recent increases were observed at Facility 117, Facility 126, Facility 26, Facility 69, and Facility 30 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). In parallel, analysis of injectable-associated risk, based on \u003cem\u003errs\u003c/em\u003e and/or \u003cem\u003eeis\u003c/em\u003e mutation proxies, identified Facility 9, Facility 117, Facility 21, Facility 138, and Facility 11 as facilities with the largest recent increases (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThese temporal deviations are visualized using quarterly time-series plots, with dashed vertical lines marking the two most recent quarters, enabling direct comparison of recent changes relative to prior facility-specific baselines. The concentration of increases within a limited number of facilities supports the use of facility-level mutation surveillance as an early-warning mechanism to flag settings with emerging resistance-associated signals.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003e \u003cb\u003eFacility-level mutation fingerprints as indicators of local selective pressure\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe facility-level mutation fingerprints identified in this study reveal substantial heterogeneity in the mutational landscape of \u003cem\u003eMycobacterium tuberculosis\u003c/em\u003e across health-care settings. Facilities clustered into distinct groups characterized by predominance of \u003cem\u003ekatG\u003c/em\u003e-associated, \u003cem\u003einhA\u003c/em\u003e-associated, or mixed isoniazid-related mutation profiles, with a smaller subset exhibiting elevated prevalence of \u003cem\u003egyrA\u003c/em\u003e-associated mutation proxies linked to fluoroquinolone resistance. This heterogeneity indicates that resistance emergence is not uniformly distributed across the health system, but instead reflects localised selective pressures shaped by treatment practices, referral pathways, and patient case mix [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Likewise, WGS-based work from high-burden settings routinely identifies katG and inhA as major INH resistance loci, but with variable proportions across study populations and catchment areas [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. A study carried out in the Western Cape, South Africa, demonstrated strong geographic clustering of ofloxacin/amikacin resistance among RR-TB, illustrating how second-line resistance can be localized rather than uniform, an epidemiologic pattern that supports our findings, facility-level gyrA subset signal [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Another study using whole genomic sequencing (WGS) similarly shows that fluoroquinolone resistance mutations (including gyrA) can be concentrated within specific transmission clusters, reinforcing the notion of localized emergence and spread [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Although our study used mutation-proxy signals (melt-peak\u0026ndash;derived profiles) rather than WGS/LPA calls; however, the pattern of localized heterogeneity is consistent with the clustering repeatedly documented by genomic epidemiology [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFacilities with \u003cem\u003ekatG\u003c/em\u003e-dominant mutation profiles likely represent settings with established or recurrent isoniazid resistance, whereas \u003cem\u003einhA\u003c/em\u003e-dominant or mixed profiles may reflect evolving resistance pathways or differential drug exposure histories [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. The identification of facilities with emerging \u003cem\u003egyrA\u003c/em\u003e-associated mutation fingerprints is of particular concern, as fluoroquinolone resistance substantially compromises the effectiveness of second-line treatment and signals progression towards more complex forms of drug-resistant TB [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eInjectable-associated mutation proxies were concentrated in a limited number of facilities and demonstrated recent upward trends in specific settings. This is consistent with evidence that SLID resistance is largely driven by recurrent high-confidence variants (notably in rrs and eis) that can cluster within specific referral or transmission networks, even as national policy shifts toward all-oral DR-TB regimens reduce injectable use overall [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. In a large South African genomic analysis of XDR-TB (Western Cape), SLID resistance was largely driven by a small set of recurrent variants in rrs and eis, with rrs 1401A\u0026thinsp;\u0026gt;\u0026thinsp;G being the dominant mutation pattern across lineages [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. This clustering suggests that early emergence of second-line resistance may be driven predominantly by facility-level amplification rather than widespread community transmission. This happens especially where there is a higher concentration of patients with prior treatment, complex DR-TB, interruptions in care, or regimen changes. Large-scale analyses that distinguish baseline versus acquired resistance highlight that resistance to newer/second-line drugs can be acquired under treatment pressure, consistent with amplification dynamics [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. In contrast, clustering does not automatically imply amplification. A study conducted in KwaZulu-Natal reported that the majority of bedaquiline-resistant cases were attributable to direct transmission of diverse resistant strains, indicating that resistance signals can become concentrated through transmission rather than through in-facility acquisition alone [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. From a public health perspective, these findings demonstrate that routinely generated molecular diagnostic data can be repurposed to produce actionable, facility-specific surveillance intelligence, enabling targeted intervention before resistance patterns become entrenched at district or provincial levels.\u003c/p\u003e \u003cp\u003eFacility-level mutation fingerprints provide a pragmatic bridge between individual-level diagnostics and population-level TB control, offering a scalable framework for early-warning surveillance using data already embedded within routine laboratory systems.\u003c/p\u003e\n\u003ch3\u003eEarly-warning signals for emerging resistance\u003c/h3\u003e\n\u003cp\u003eOur study demonstrates the feasibility and public health value of using facility-level temporal changes in mutation proxy prevalence as early warning signals for emerging drug resistance. By focusing on quarter-to-quarter deviations rather than absolute prevalence alone, the analysis identifies facilities where resistance-associated signals are accelerating relative to their historical baselines, a pattern particularly relevant for anticipatory surveillance.\u003c/p\u003e \u003cp\u003eThe detection of pronounced recent increases in gyrA mutation proxies at a small subset of facilities (Facilities 117, 126, 26, 69, and 30) is epidemiologically significant. Fluoroquinolone resistance represents a critical inflection point in the drug-resistance continuum because fluoroquinolones are core Group A drugs in second-line regimens, and resistance to any fluoroquinolone defines pre-XDR-TB under WHO definitions [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. In recent programmatic studies, fluoroquinolone resistance has remained associated with poorer treatment outcomes compared with MDR/RR-TB without FQ resistance, although the magnitude of this effect may vary with access to effective all-oral regimens and timely regimen optimisation [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. The fact that these increases are concentrated within a limited number of facilities suggests localized amplification or selective pressure, rather than widespread community-level dissemination. The facility-level clustering and recent increases are consistent with localized amplification or selective pressure linked to treatment and prescribing pathways; however, recent WGS studies indicate that second-line resistance can also be driven by transmission within connected referral and community networks, which may present as geographically or facility-associated clustering. Further linkage of mutation-proxy trends to treatment history (baseline vs follow-up) and genomic clustering would help distinguish amplification from transmission in this setting [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. Such patterns are unlikely to be detected through conventional aggregate reporting, which typically smooths short-term fluctuations and dilutes facility-specific signals.\u003c/p\u003e \u003cp\u003eSimilarly, the identification of recent increases in injectable-associated mutation proxies (rrs and/or eis) at a distinct but overlapping set of facilities (Facilities 9, 117, 21, 138, and 11) underscores the capacity of facility-level monitoring to detect early shifts in second-line resistance risk. This finding is consistent with evidence from molecular epidemiology that second-line drug resistance is not evenly distributed but instead may cluster spatially or programmatically where selective pressures or transmission networks are concentrated [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. This supports the idea that facility or district characteristics can shape where resistant variants first emerge and are most detectable. Such heterogeneity underscores the potential value of facility-level mutation monitoring as an early-warning mechanism that can detect emerging resistance trends before they become more common at larger geographic scales.\u003c/p\u003e \u003cp\u003eHowever, it is also important to recognise that clustering of resistance does not exclusively indicate de novo amplification; in some high-burden settings, genomic analyses have shown that transmission of resistant strains contributes significantly to observed resistance patterns, which may still appear geographically localized depending on connected transmission networks across facilities and communities [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Distinguishing between amplification and transmission pathways may require linking proxy mutation data to patient treatment histories and performing fine-scale genomic clustering. Although injectable resistance remains relatively uncommon overall, its emergence in specific facilities may reflect treatment-history effects, referral of complex cases, or programmatic challenges such as delayed regimen modification or suboptimal adherence support. Recent laboratory and programmatic studies show that SLID resistance is typically detected at low frequency when broader DR-TB samples are analysed, but clusters appear in higher-risk subgroups. Similarly, multicountry evaluations of Xpert MTB/XDR in sub-Saharan Africa demonstrate that detection of aminoglycoside/injectable resistance targets (rrs/eis-related) occurs in a minority of tested DR-TB cases, reflecting the relatively lower prevalence of these mutations compared with INH- or FQ-associated loci in many routine cohorts [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Importantly, Facility 117 appeared in both fluoroquinolone- and injectable-associated early-warning analyses, highlighting how convergent resistance signals across drug classes can flag facilities at particularly high risk of progression toward extensively drug-resistant TB.\u003c/p\u003e \u003cp\u003eThe use of quarterly time-series plots with facility-specific baselines provides a critical interpretive advantage. By visualizing recent changes alongside historical trends, this approach distinguishes meaningful deviations from expected background variability and avoids over-interpretation of single-period spikes. The dashed demarcation of the two most recent quarters facilitates rapid appraisal by programme managers, enabling translation of complex molecular data into an intuitive governance signal. This design aligns with the concept of \u003cem\u003esignal-based surveillance\u003c/em\u003e, where the timing and direction of change are as informative as absolute prevalence [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFrom a surveillance perspective, these findings support the argument that facility-level mutation surveillance can function as a leading indicator of resistance emergence, potentially preceding formal increases in categorical resistance interpretations [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. Recent genomic surveillance studies similarly emphasize early detection of resistance mutations to guide timely programmatic action, and implementation studies of second-line reflex testing demonstrate the feasibility of embedding such early-warning approaches in routine laboratory workflows [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. Because mutation proxies capture underlying genetic changes directly, they may provide earlier insight into evolving resistance dynamics than phenotype-based summaries, which often lag due to diagnostic, reporting, and aggregation delays. In high-burden settings with constrained access to routine sequencing, this approach offers a pragmatic alternative for strengthening resistance intelligence using existing diagnostic infrastructure.\u003c/p\u003e\n\u003ch3\u003ePolicy and programmatic implications\u003c/h3\u003e\n\u003cp\u003eThe facility-level mutation fingerprinting framework presented in this study is directly relevant to national and subnational TB control programs. By leveraging routinely generated laboratory molecular diagnostic outputs, this approach enables near\u0026ndash;real-time identification of facilities exhibiting emerging drug-resistance signals, without additional laboratory testing or resource investment. Integration of mutation fingerprint surveillance into existing TB monitoring dashboards could support targeted programmatic actions, including intensified adherence support, regimen review, enhanced infection prevention and control, and focused clinical governance at facilities flagged through early-warning indicators. At a policy level, this framework aligns with precision public health principles by shifting drug-resistance surveillance from passive reporting to proactive, data-driven intervention, and could complement existing phenotypic and sequencing-based surveillance systems. Adoption of this approach has the potential to strengthen TB drug-resistance containment strategies, optimize resource allocation, and improve treatment outcomes in high-burden settings.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study demonstrates that routine molecular diagnostic data can be repurposed beyond individual-level resistance reporting to support facility-level genomic surveillance of drug-resistant \u003cem\u003eMycobacterium tuberculosis\u003c/em\u003e. By operationalizing mutant melt peak temperatures as mutation-proxy signals and aggregating these data across time and facilities, we identified distinct facility mutation fingerprints, strong associations between mutation proxies and phenotypic resistance, and early-warning signals indicative of emerging resistance risk.\u003c/p\u003e \u003cp\u003eThe findings highlight substantial heterogeneity in mutational patterns across facilities, suggesting that resistance emergence is shaped by local selective pressures rather than occurring uniformly across the health system. The strong, near-deterministic relationships observed between key mutation proxies and resistance interpretations underscore the robustness of routine molecular outputs as indicators of underlying genetic resistance mechanisms. Importantly, the identification of recent facility-level increases in fluoroquinolone- and injectable-associated mutation proxies illustrates the potential of this approach to detect incipient resistance trends earlier than conventional summary reporting.\u003c/p\u003e \u003cp\u003eFrom a public health perspective, facility-level mutation fingerprinting offers a practical, scalable, and cost-neutral surveillance framework that leverages existing laboratory infrastructure without requiring additional testing or sequencing capacity. Integrating such analyses into routine TB monitoring systems could enable more timely, targeted programmatic responses, including focused adherence support, treatment review, and infection control interventions at facilities exhibiting early warning signals.\u003c/p\u003e \u003cp\u003eFuture work should explore integrating mutation fingerprint surveillance with patient-level treatment histories, referral pathways, and programmatic interventions, and assess its applicability in other high-burden settings. Overall, this study provides proof of concept for transforming routine molecular diagnostic data into actionable intelligence to strengthen TB drug-resistance surveillance and control.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003e\u003cstrong\u003eAI:\u003c/strong\u003e Artificial intelligence\u003cbr\u003e\u003cstrong\u003eAMK:\u003c/strong\u003e Amikacin\u003cbr\u003e\u003cstrong\u003eCAP:\u003c/strong\u003e Capreomycin\u003cbr\u003e\u003cstrong\u003eCG:\u003c/strong\u003e Clinical Governance\u003cbr\u003e\u003cstrong\u003eCI:\u003c/strong\u003e Confidence interval\u003cbr\u003e\u003cstrong\u003eCXR:\u003c/strong\u003e Chest X-ray\u003cbr\u003e\u003cstrong\u003eDR-TB:\u003c/strong\u003e Drug-resistant tuberculosis\u003cbr\u003e\u003cstrong\u003eeis:\u003c/strong\u003e Enhanced intracellular survival gene\u003cbr\u003e\u003cstrong\u003eETH:\u003c/strong\u003e Ethionamide\u003cbr\u003e\u003cstrong\u003eFQ / FLQ:\u003c/strong\u003e Fluoroquinolones\u003cbr\u003e\u003cstrong\u003eHREC:\u003c/strong\u003e Health Research Ethics Committee\u003cbr\u003e\u003cstrong\u003eINH:\u003c/strong\u003e Isoniazid\u003cbr\u003e\u003cstrong\u003eKAN:\u003c/strong\u003e Kanamycin\u003cbr\u003e\u003cstrong\u003ekatG:\u003c/strong\u003e Catalase-peroxidase gene associated with isoniazid resistance\u003cbr\u003e\u003cstrong\u003eLPA:\u003c/strong\u003e Line probe assay\u003cbr\u003e\u003cstrong\u003eMDR-TB:\u003c/strong\u003e Multidrug-resistant tuberculosis\u003cbr\u003e\u003cstrong\u003eMTB:\u003c/strong\u003e \u003cem\u003eMycobacterium tuberculosis\u003c/em\u003e\u003cbr\u003e\u003cstrong\u003eNHLS:\u003c/strong\u003e National Health Laboratory Service\u003cbr\u003e\u003cstrong\u003eOR:\u003c/strong\u003e Odds ratio\u003cbr\u003e\u003cstrong\u003ePCR:\u003c/strong\u003e Polymerase chain reaction\u003cbr\u003e\u003cstrong\u003ePTM:\u003c/strong\u003e Pretomanid\u003cbr\u003e\u003cstrong\u003erpoB:\u003c/strong\u003e RNA polymerase \u0026beta;-subunit gene\u003cbr\u003e\u003cstrong\u003errs:\u003c/strong\u003e 16S ribosomal RNA gene associated with aminoglycoside resistance\u003cbr\u003e\u003cstrong\u003eSLID:\u003c/strong\u003e Second-line injectable drug\u003cbr\u003e\u003cstrong\u003eTB:\u003c/strong\u003e Tuberculosis\u003cbr\u003e\u003cstrong\u003eWGS:\u003c/strong\u003e Whole-genome sequencing\u003cbr\u003e\u003cstrong\u003eWSU:\u003c/strong\u003e Walter Sisulu University\u003cbr\u003e\u003cstrong\u003eXDR-TB:\u003c/strong\u003e Extensively drug-resistant tuberculosis\u003cbr\u003e\u003cstrong\u003eXpert MTB/XDR:\u003c/strong\u003e Molecular diagnostic assay for resistance detection\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study complied with the ethical principles of the Declaration of Helsinki. Ethical approval was obtained from the Research Ethics Biosafety Committee, Faculty of Health Sciences, Walter Sisulu University (WSU HREC 141/2025; approved July 2, 2025). The Eastern Cape Department of Health granted administrative authorization to access and analyse facility-level data (Ref: EC_202507_023; approved July 11, 2025). The National Health Laboratory Service approved the use of de-identified laboratory diagnostic data for research purposes (SR4169693; approved November 17, 2025). Because the study used de-identified secondary data generated during routine diagnostic services, with no direct patient contact or additional specimen collection, the ethics committee approved the study and waived the requirement for individual informed consent.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable. This study uses de-identified secondary data with no individual-level identifiable information.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial registration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eClinical trial number: \u003cstrong\u003eNot applicable.\u003c/strong\u003e This study did not involve an interventional clinical trial and, therefore, did not require trial registration.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and analyzed during the current study are not publicly available due to data protection regulations of the Eastern Cape Department of Health and the National Health Laboratory Service. De-identified aggregated data may be made available from the corresponding author upon reasonable request and with permission of the respective data custodians.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors are pleased to share that they have no competing interests to disclose that relate to the content of this article.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Institutional support was provided by Walter Sisulu University, Faculty of Health Sciences.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLindiwe Modest Faye\u003c/strong\u003e: Conceptualization, methodology, supervision, formal analysis, data interpretation, original draft writing, review, and editing. \u003cstrong\u003eMelisa Makhuba:\u003c/strong\u003e Investigation, data curation, laboratory work, resources, formal analysis, review, and editing. \u003cstrong\u003eNtandazo Dlatu\u003c/strong\u003e: Software, statistical analysis, validation, visualization, review, and editing. \u003cstrong\u003eMojisola Clara Hosu:\u003c/strong\u003e Literature review, methodology, project administration, data interpretation, review, and editing. \u003cstrong\u003eTeke Apalata\u003c/strong\u003e: Supervision, clinical and laboratory oversight, funding acquisition, project administration, review and editing, and final approval of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors acknowledge the Eastern Cape Department of Health, the National Health Laboratory Service, and the participating health facilities for their cooperation. We also thank the Walter Sisulu University TB Research Unit for technical and administrative support.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAuld SC, Shah NS, Mathema B, Brown TS, Ismail N, Omar SV, Brust JCM, Nelson KN, Allana S, Campbell A, Mlisana K. Extensively drug-resistant tuberculosis in South Africa: genomic evidence supporting transmission in communities. Eur Respir J. 2018;52(4):1800246. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1183/13993003.00246-2018\u003c/span\u003e\u003cspan address=\"10.1183/13993003.00246-2018\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNaidoo K, Perumal R, Cox H, Mathema B, Loveday M, Ismail N, Omar SV, Georghiou SB, Daftary A, O'Donnell M, Ndjeka N. The epidemiology, transmission, diagnosis, and management of drug-resistant tuberculosis: lessons from the South African experience. Lancet Infect Dis. 2024;24(9):e559\u0026ndash;75. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S1473-3099(24)00144-0\u003c/span\u003e\u003cspan address=\"10.1016/S1473-3099(24)00144-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElbehiry A, Marzouk E, Edrees HM, AlShaqi R, Ellethy AT, Alzaben F, Anagreyyah S, Algarni A, Almuhaydili K, Alotaibi I, Albaqami A. Advancing the fight against tuberculosis: integrating innovation and public health in diagnosis, treatment, vaccine development, and implementation science. Front Med. 2025;12:1596579. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fmed.2025.1596579\u003c/span\u003e\u003cspan address=\"10.3389/fmed.2025.1596579\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUllah MN, Khan HR, Imran A, Saeed A, Yasin U, Ahmed N, Shah SAH. The future of Mycobacterium tuberculosis diagnostics: integrating mass spectrometry, AI, and genomics. J Health Wellness Community Res. 2025;e334. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.61919/0fvx1h30\u003c/span\u003e\u003cspan address=\"10.61919/0fvx1h30\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao Y, Parmar H, Simmons AM, Kale D, Tong K, Lieu D, Persing D, Kwiatkowski R, Alland D, Chakravorty S. Automatic identification of rpoB mutations responsible for rifampin resistance using melting temperature signatures with Xpert MTB/RIF Ultra. J Clin Microbiol. 2019;58(1):10\u0026ndash;1128. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1128/jcm.00907-19\u003c/span\u003e\u003cspan address=\"10.1128/jcm.00907-19\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTorrea G, Ng KC, Van Deun A, Andr\u0026eacute; E, Kaisergruber J, Ssengooba W, Desmaretz C, Gabriels S, Driesen M, Diels M, Asnong S. Variable ability of rapid tests to detect Mycobacterium tuberculosis rpoB mutations conferring phenotypically occult rifampicin resistance. Sci Rep. 2019;9:11826. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-019-48401-z\u003c/span\u003e\u003cspan address=\"10.1038/s41598-019-48401-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarcia LS, Costa AG, Ara\u0026uacute;jo-Pereira M, Spener-Gomes R, Fran\u0026ccedil;a Aguiar A, Souza AB, Lima LO, Benjamin A, Rocha MS, Moreira AS, Silva J. Xpert MTB/RIF cycle threshold predicts TB transmission to close contacts in Brazil. Clin Infect Dis. 2024;ciad794.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo Y, Yang J, Wang H, Sha W, Yu F. Key resistance-associated mutations in MDR-TB with focus on aminoglycosides: a genomic study from Shanghai, China. BMC Microbiol. 2025;25:702. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12866-025-04446-x\u003c/span\u003e\u003cspan address=\"10.1186/s12866-025-04446-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSingh AK, Singh N, Kumar S, Mishra AK, Singh NP. Molecular insights of drug-resistant tuberculosis: genetic mutations and their profile. Front Microbiol. 2025;16:1669327. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fmicb.2025.1669327\u003c/span\u003e\u003cspan address=\"10.3389/fmicb.2025.1669327\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGilmour B, Alene KA. Ending tuberculosis: challenges and opportunities. Front Tuberc. 2024;2:1487518. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/ftubr.2024.1487518\u003c/span\u003e\u003cspan address=\"10.3389/ftubr.2024.1487518\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDing F, Liu W, Wu C, Zhang W, Chen S, Lai W, Qu J, Lin Q, Lu S, Qu J. Whole-genome sequencing reveals TB transmission patterns and drug resistance within/between hosts. Front Cell Infect Microbiol. 2025;14:1488547. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fcimb.2024.1488547\u003c/span\u003e\u003cspan address=\"10.3389/fcimb.2024.1488547\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGopalaswamy R, Subbian S. The power of resistance: mechanisms of antimicrobial resistance in Mycobacterium tuberculosis and impact on TB management. Clin Microbiol Rev. 2026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1128/cmr.00194-25\u003c/span\u003e\u003cspan address=\"10.1128/cmr.00194-25\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. e00194-25.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Zwaan R, de Vries G, Ubbelohde E, Mulder A, Kamst-van Agterveld M, Rebel K, Kautz S, Kremer K, Anthony RM, van Soolingen D. Verification of emerging genomic mutations enables distinguishing TB transmission chains over 30 years. PLoS ONE. 2025;20(6):e0319630. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0319630\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0319630\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarcon DJ, Sharma A, Souza AB, Barros RB, Andrade VDGD, Guimar\u0026atilde;es RJDPS, Lima LNG, Monteiro LHMT, Quaresma AJPG, Ribeiro LR, Suffys PN. Genomic surveillance reveals XDR-TB transmission profiles in Par\u0026aacute;, Brazil. Front Microbiol. 2025;15:1514862. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fmicb.2024.1514862\u003c/span\u003e\u003cspan address=\"10.3389/fmicb.2024.1514862\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNgom JT, Loubser J, Maasdorp E, Ghebrekristos Y, Singh S, Opperman CJ, Klopper M, Warren RM, Streicher EM. Population structure \u0026amp; emerging resistance to new TB drugs in Western Cape: 10-year genomic study. Front Cell Infect Microbiol. 2025;15:1638577. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fcimb.2025.1638577\u003c/span\u003e\u003cspan address=\"10.3389/fcimb.2025.1638577\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNono VN, Nantia EA, Mutshembele A, Teagho SN, Simo YWK, Takong BS, Djieugoue YJ, Assolo YP, Ongboulal SM, Awungafac SN, Eyangoh S. Prevalence of katG and inhA mutations linked to isoniazid resistance in Cameroon. BMC Microbiol. 2025;25(1):127. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12866-025-03816-9\u003c/span\u003e\u003cspan address=\"10.1186/s12866-025-03816-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePei S, Song Z, Yang W, He W, Ou X, Zhao B, He P, Zhou Y, Xia H, Wang S, Jia Z. Catalogue of M. tuberculosis mutations associated with drug resistance to 12 drugs in China: nationwide genomic study. Lancet Microbe. 2024;5(11):100899. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S2666-5247(24)00131-9\u003c/span\u003e\u003cspan address=\"10.1016/S2666-5247(24)00131-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSy KTL, Leavitt SV, de Vos M, Dolby T, Bor J, Horsburgh CR Jr, Warren RM, Streicher EM, Jenkins HE, Jacobson KR. Spatial heterogeneity of XDR-TB in Western Cape Province, South Africa. Sci Rep. 2022;12:10844. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-022-14581-4\u003c/span\u003e\u003cspan address=\"10.1038/s41598-022-14581-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiriba G, Mollalign H, Meaza A, Getu M, Zerihun B, Alemu A, Dagne B, Abebaw Y, Weldemariam AG, Bashea C, Ali A. WGS-based detection of XDR-TB in Ethiopia. Commun Med. 2026;6:14. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s43856-025-01271-1\u003c/span\u003e\u003cspan address=\"10.1038/s43856-025-01271-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eValafar SJ. Systematic review of mutations associated with isoniazid resistance points to continuing evolution and subsequent evasion of molecular detection, and potential for emergence of multidrug resistance in clinical strains of Mycobacterium tuberculosis. Antimicrob Agents Chemother. 2021;65(3):10\u0026ndash;128.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNgema SL, Dookie N, Perumal R, Nandlal L, Naicker N, Letsoalo MP, O'Donnell M, Khan A, Padayatchi N, Naidoo K. Isoniazid resistance-conferring mutations are associated with highly variable phenotypic resistance. J Clin Tuberc Other Mycobact Dis. 2023;33:100387.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDean AS, Auguet OT, Glaziou P, Zignol M, Ismail N, Kasaeva T, Floyd K. 25 years of surveillance of drug-resistant tuberculosis: achievements, challenges, and way forward. Lancet Infect Dis. 2022;22(7):e191\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiebenberg D, Gordhan BG, Kana BD. Drug-resistant tuberculosis: implications for transmission, diagnosis, and disease management. Front Cell Infect Microbiol. 2022;12:943545.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiao YX, Liu KH, Lin WH, Chan TH, Jou R. Whole-genome sequencing-based analyses of drug-resistant Mycobacterium tuberculosis from Taiwan. Sci Rep. 2023;13:2540. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-023-29652-3\u003c/span\u003e\u003cspan address=\"10.1038/s41598-023-29652-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOliveira RS, Brandao AP, Ferreira FMDA, Costa SMD, Silva VLM, Ferrazoli L, Chimara E, Pinhata JMW. Fluoroquinolone and second-line injectable resistance among RIF- and INH-resistant M. tuberculosis. Microorganisms. 2025;13:2470. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/microorganisms13112470\u003c/span\u003e\u003cspan address=\"10.3390/microorganisms13112470\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTimm J, Bateson A, Solanki P, Paleckyte A, Witney AA, Rofael SA, Fabiane S, Olugbosi M, McHugh TD, Sun E. Baseline and acquired resistance to BDQ, LZD and PTM in four clinical trials. PLOS Glob Public Health. 2023;3:e0002283. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pgph.0002283\u003c/span\u003e\u003cspan address=\"10.1371/journal.pgph.0002283\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMboowa G. Reimagining TB control in the genomics era: global investment in TB genomic surveillance. Pathogens. 2025;14:975. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/pathogens14100975\u003c/span\u003e\u003cspan address=\"10.3390/pathogens14100975\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSobol R, Omar SV, Brown T, Joseph L, Lutchminarian K, Tang L, Lan Y, Willis F, Campbell A, Warren JL, Cohen T. Transmission of BDQ-resistant TB in KwaZulu-Natal. Am J Respir Crit Care Med. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1164/rccm.202506-1489OC\u003c/span\u003e\u003cspan address=\"10.1164/rccm.202506-1489OC\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization. WHO announces updated definitions of extensively drug-resistant tuberculosis. 2021. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.who.int/news/item/27-01-2021-who-announces-updated-definitions-of-extensively-drug-resistant-tuberculosis\u003c/span\u003e\u003cspan address=\"https://www.who.int/news/item/27-01-2021-who-announces-updated-definitions-of-extensively-drug-resistant-tuberculosis\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNehru VJ, Jose Vandakunnel M, Brammacharry U, Ramachandra V, Pradhabane G, Mani BR, Vn AD, Muthaiah M. FQ resistance transmission in DR-TB: genomic epidemiology study. Sci Rep. 2024;14:19719. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-024-70535-y\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-70535-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim S, Mok J. Treatment outcomes of FQ-resistant MDR-TB: implications for delamanid\u0026mdash;authors\u0026rsquo; reply. Tuberc Respir Dis. 2024;87:209\u0026ndash;11. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4046/trd.2024.0010\u003c/span\u003e\u003cspan address=\"10.4046/trd.2024.0010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi M, Zhang Y, Wu Z, Jiang Y, Sun R, Yang J, Li J, Lin H, Zhang R, Jiang Q, Wang L. Transmission of FQ resistance in MDR-TB in Shanghai: WGS epidemiology. Emerg Microbes Infect. 2024;13:2302837. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/22221751.2024.2302837\u003c/span\u003e\u003cspan address=\"10.1080/22221751.2024.2302837\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Y, Li D, Liu J, Jiang Y, Shen X, Xu B. Pre-diagnostic FQ exposure \u0026amp; acquired resistance in M. tuberculosis: WGS-based case-control study. J Infect. 2025;91:106648.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMassou F, Diarra B, Diallo AB, Bah KS, Vuchas C, Neh A, Sanders M, Adebiyi EO, Aderemi BO, Agbla SC, Flyod S. Multicenter evaluation of Xpert MTB/XDR in Sub-Saharan Africa. ERJ Open Res. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1183/23120541.00427-2025\u003c/span\u003e\u003cspan address=\"10.1183/23120541.00427-2025\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRilkoff H, Struck S, Ziegler C, Faye L, Paquette D, Buckeridge D. Innovations in public health surveillance: novel data \u0026amp; analytic methods. Can Commun Dis Rep. 2024;50:93\u0026ndash;101. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.14745/ccdr.v50i34a02\u003c/span\u003e\u003cspan address=\"10.14745/ccdr.v50i34a02\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHicketier A, Bach M, Oedi P, Ullrich A, Abbood A. Ensemble-labeling of infectious disease time series to evaluate early warning systems. Infect Dis Model. 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarilar I, Fernando T, Utpatel C, Abujate C, Madeira CM, Jos\u0026eacute; B, Mutaquiha C, Kranzer K, Niemann T, Ismael N, de Araujo L. Emergence of BDQ-resistant TB and undetected rpoB Ile491Phe MDR/XDR-TB in Mozambique. Lancet Infect Dis. 2024;24:297\u0026ndash;307. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S1473-3099(23)00498-X\u003c/span\u003e\u003cspan address=\"10.1016/S1473-3099(23)00498-X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMakondo VT, Kaapu KG, Wells F, Sharma A, Lekalakala-Mokaba MR, Warren R, Concei\u0026ccedil;\u0026atilde;o EC, Rukasha I. Empowering TB genomic surveillance in Limpopo, South Africa. Front Public Health. 2025;13:1567382. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fpubh.2025.1567382\u003c/span\u003e\u003cspan address=\"10.3389/fpubh.2025.1567382\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCentner CM, Munir R, Tagliani E, Rie\u0026szlig; F, Brown P, Hayes C, Dolby T, Zemanay W, Cirillo DM, David A, Schumacher SG. Reflex Xpert MTB/XDR testing of RR-TB specimens: diagnostic accuracy and feasibility. Open Forum Infect Dis. 2024;11:ofae437. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/ofid/ofae437\u003c/span\u003e\u003cspan address=\"10.1093/ofid/ofae437\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-infectious-diseases","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"infd","sideBox":"Learn more about [BMC Infectious Diseases](http://bmcinfectdis.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/infd","title":"BMC Infectious Diseases","twitterHandle":"#bmcinfectdis","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Mycobacterium tuberculosis, drug resistance, molecular surveillance, mutation proxies, facility-level analysis, early-warning systems","lastPublishedDoi":"10.21203/rs.3.rs-8790685/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8790685/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eRoutine molecular tuberculosis (TB) diagnostics generate high-dimensional \"off-protocol\" data, including mutant melt peak temperatures and cycle threshold (CT) values. These data are currently underutilized, typically discarded after individual resistance reporting. We aimed to evaluate whether aggregating these routine \"mutation-proxy\" signals could provide a scalable framework for facility-level surveillance and early warning of emerging drug resistance.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eWe conducted a retrospective longitudinal study of 4,300 TB diagnostic episodes across 139 health-care facilities over 16 quarters. Mutation-proxy signals for five key loci (\u003cem\u003ekatG, inhA, gyrA, rrs, eis\u003c/em\u003e) were extracted from raw diagnostic outputs. We constructed \"facility-level mutation fingerprints\" by aggregating prevalence data and employed hierarchical clustering to identify distinct resistance topographies. Associations between proxies and laboratory-confirmed resistance were modelled using L2-regularized (ridge) logistic regression with facility-level cluster bootstrap confidence intervals to account for near-separation and spatial autocorrelation.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eIsoniazid-associated proxies predominated (\u003cem\u003ekatG\u003c/em\u003e: 46.3%; \u003cem\u003einhA\u003c/em\u003e: 25.1%), while \u003cem\u003egyrA\u003c/em\u003e (fluoroquinolone-associated) and \u003cem\u003errs\u003c/em\u003e (injectable-associated) proxies were detected in 12.5% and 7.7% of episodes, respectively. Clustering revealed four distinct facility profiles: \u003cem\u003ekatG\u003c/em\u003e-dominant, \u003cem\u003einhA\u003c/em\u003e-dominant, mixed-isoniazid, and a high-risk \"emerging \u003cem\u003egyrA\u003c/em\u003e\" profile. Regression analysis confirmed high diagnostic accuracy for the proxies, notably for isoniazid (katG: OR\u0026thinsp;=\u0026thinsp;1,146; inhA: OR\u0026thinsp;=\u0026thinsp;603) and fluoroquinolones (gyrA: OR\u0026thinsp;=\u0026thinsp;7,136). Longitudinal analysis successfully identified a subset of facilities that exhibited significant quarter-over-quarter increases in second-line resistance proxies prior to traditional surveillance detection.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eFacility-level mutation fingerprinting leverages existing, \"near-zero-cost\" laboratory data to provide a granular, real-time map of the resistance landscape. This framework enables precision public health interventions, allowing TB programmes to transition from reactive to proactive, facility-targeted containment of emerging drug-resistant \u003cem\u003eMycobacterium tuberculosis\u003c/em\u003e.\u003c/p\u003e","manuscriptTitle":"Facility-Level Mutation Fingerprints and Early-Warning Surveillance of Drug-Resistant Mycobacterium tuberculosis in Rural Eastern Cape","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-25 18:29:27","doi":"10.21203/rs.3.rs-8790685/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"242215192486596947116294810684423853966","date":"2026-02-26T07:29:29+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-20T16:26:00+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-02-11T04:16:05+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-09T00:14:37+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-09T00:13:44+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Infectious Diseases","date":"2026-02-04T23:01:38+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-infectious-diseases","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"infd","sideBox":"Learn more about [BMC Infectious Diseases](http://bmcinfectdis.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/infd","title":"BMC Infectious Diseases","twitterHandle":"#bmcinfectdis","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"87781678-2e53-4950-973a-71282af0be31","owner":[],"postedDate":"February 25th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-02-25T18:29:28+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-25 18:29:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8790685","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8790685","identity":"rs-8790685","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}